1 Document: draft-cheshire-dnsext-multicastdns-06.txt Stuart Cheshire
2 Internet-Draft Marc Krochmal
3 Category: Standards Track Apple Computer, Inc.
4 Expires 10th February 2007 10th August 2006
8 <draft-cheshire-dnsext-multicastdns-06.txt>
12 By submitting this Internet-Draft, each author represents that any
13 applicable patent or other IPR claims of which he or she is aware
14 have been or will be disclosed, and any of which he or she becomes
15 aware will be disclosed, in accordance with Section 6 of BCP 79.
16 For the purposes of this document, the term "BCP 79" refers
17 exclusively to RFC 3979, "Intellectual Property Rights in IETF
18 Technology", published March 2005.
20 Internet-Drafts are working documents of the Internet Engineering
21 Task Force (IETF), its areas, and its working groups. Note that
22 other groups may also distribute working documents as Internet-
25 Internet-Drafts are draft documents valid for a maximum of six months
26 and may be updated, replaced, or obsoleted by other documents at any
27 time. It is inappropriate to use Internet-Drafts as reference
28 material or to cite them other than as "work in progress."
30 The list of current Internet-Drafts can be accessed at
31 http://www.ietf.org/1id-abstracts.html
33 The list of Internet-Draft Shadow Directories can be accessed at
34 http://www.ietf.org/shadow.html
38 As networked devices become smaller, more portable, and
39 more ubiquitous, the ability to operate with less configured
40 infrastructure is increasingly important. In particular,
41 the ability to look up DNS resource record data types
42 (including, but not limited to, host names) in the absence
43 of a conventional managed DNS server, is becoming essential.
45 Multicast DNS (mDNS) provides the ability to do DNS-like operations
46 on the local link in the absence of any conventional unicast DNS
47 server. In addition, mDNS designates a portion of the DNS namespace
48 to be free for local use, without the need to pay any annual fee, and
49 without the need to set up delegations or otherwise configure a
50 conventional DNS server to answer for those names.
52 The primary benefits of mDNS names are that (i) they require little
53 or no administration or configuration to set them up, (ii) they work
54 when no infrastructure is present, and (iii) they work during
55 infrastructure failures.
58 Expires 10th February 2007 Cheshire & Krochmal [Page 1]
60 Internet Draft Multicast DNS 10th August 2006
65 1. Introduction....................................................3
66 2. Conventions and Terminology Used in this Document...............3
67 3. Multicast DNS Names.............................................4
68 4. Source Address Check............................................8
69 5. Reverse Address Mapping.........................................9
70 6. Querying.......................................................10
71 7. Duplicate Suppression..........................................15
72 8. Responding.....................................................17
73 9. Probing and Announcing on Startup..............................20
74 10. Conflict Resolution............................................26
75 11. Resource Record TTL Values and Cache Coherency.................28
76 12. Special Characteristics of Multicast DNS Domains...............33
77 13. Multicast DNS for Service Discovery............................34
78 14. Enabling and Disabling Multicast DNS...........................34
79 15. Considerations for Multiple Interfaces.........................35
80 16. Considerations for Multiple Responders on the Same Machine.....36
81 17. Multicast DNS and Power Management.............................38
82 18. Multicast DNS Character Set....................................39
83 19. Multicast DNS Message Size.....................................41
84 20. Multicast DNS Message Format...................................42
85 21. Choice of UDP Port Number......................................45
86 22. Summary of Differences Between Multicast DNS and Unicast DNS...46
87 23. Benefits of Multicast Responses................................47
88 24. IPv6 Considerations............................................48
89 25. Security Considerations........................................49
90 26. IANA Considerations............................................50
91 27. Acknowledgments................................................50
92 28. Deployment History.............................................50
93 29. Copyright Notice...............................................51
94 30. Normative References...........................................51
95 31. Informative References.........................................52
96 32. Authors' Addresses.............................................53
116 Expires 10th February 2007 Cheshire & Krochmal [Page 2]
118 Internet Draft Multicast DNS 10th August 2006
123 When reading this document, familiarity with the concepts of Zero
124 Configuration Networking [ZC] and automatic link-local addressing
125 [RFC 2462] [RFC 3927] is helpful.
127 This document proposes no change to the structure of DNS messages,
128 and no new operation codes, response codes, or resource record types.
129 This document simply discusses what needs to happen if DNS clients
130 start sending DNS queries to a multicast address, and how a
131 collection of hosts can cooperate to collectively answer those
132 queries in a useful manner.
134 There has been discussion of how much burden Multicast DNS might
135 impose on a network. It should be remembered that whenever IPv4 hosts
136 communicate, they broadcast ARP packets on the network on a regular
137 basis, and this is not disastrous. The approximate amount of
138 multicast traffic generated by hosts making conventional use of
139 Multicast DNS is anticipated to be roughly the same order of
140 magnitude as the amount of broadcast ARP traffic those hosts already
143 New applications making new use of Multicast DNS capabilities for
144 unconventional purposes may generate more traffic. If some of those
145 new applications are "chatty", then work will be needed to help them
146 become less chatty. When performing any analysis, it is important
147 to make a distinction between the application behavior and the
148 underlying protocol behavior. If a chatty application uses UDP,
149 that doesn't mean that UDP is chatty, or that IP is chatty, or that
150 Ethernet is chatty. What it means is that the application is chatty.
151 The same applies to any future applications that may decide to layer
152 increasing portions of their functionality over Multicast DNS.
155 2. Conventions and Terminology Used in this Document
157 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
158 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
159 document are to be interpreted as described in "Key words for use in
160 RFCs to Indicate Requirement Levels" [RFC 2119].
162 This document uses the term "host name" in the strict sense to mean
163 a fully qualified domain name that has an address record. It does
164 not use the term "host name" in the commonly used but incorrect
165 sense to mean just the first DNS label of a host's fully qualified
168 A DNS (or mDNS) packet contains an IP TTL in the IP header, which
169 is effectively a hop-count limit for the packet, to guard against
170 routing loops. Each Resource Record also contains a TTL, which is
171 the number of seconds for which the Resource Record may be cached.
174 Expires 10th February 2007 Cheshire & Krochmal [Page 3]
176 Internet Draft Multicast DNS 10th August 2006
179 In any place where there may be potential confusion between these two
180 types of TTL, the term "IP TTL" is used to refer to the IP header TTL
181 (hop limit), and the term "RR TTL" is used to refer to the Resource
182 Record TTL (cache lifetime).
184 When this document uses the term "Multicast DNS", it should be taken
185 to mean: "Clients performing DNS-like queries for DNS-like resource
186 records by sending DNS-like UDP query and response packets over IP
187 Multicast to UDP port 5353."
189 This document uses the terms "shared" and "unique" when referring to
190 resource record sets.
192 A "shared" resource record set is one where several Multicast DNS
193 responders may have records with that name, rrtype, and rrclass, and
194 several responders may respond to a particular query.
196 A "unique" resource record set is one where all the records with
197 that name, rrtype, and rrclass are conceptually under the control
198 or ownership of a single responder, and it is expected that at most
199 one responder should respond to a query for that name, rrtype, and
200 rrclass. Before claiming ownership of a unique resource record set,
201 a responder MUST probe to verify that no other responder already
202 claims ownership of that set, as described in Section 9.1 "Probing".
203 For fault-tolerance and other reasons it is permitted sometimes to
204 have more than one responder answering for a particular "unique"
205 resource record set, but such cooperating responders MUST give
206 answers containing identical rdata for these records or the
207 answers will be perceived to be in conflict with each other.
209 Strictly speaking the terms "shared" and "unique" apply to resource
210 record sets, not to individual resource records, but it is sometimes
211 convenient to talk of "shared resource records" and "unique resource
212 records". When used this way, the terms should be understood to mean
213 a record that is a member of a "shared" or "unique" resource record
217 3. Multicast DNS Names
219 This document proposes that the DNS top-level domain ".local." be
220 designated a special domain with special semantics, namely that any
221 fully-qualified name ending in ".local." is link-local, and names
222 within this domain are meaningful only on the link where they
223 originate. This is analogous to IPv4 addresses in the 169.254/16
224 prefix, which are link-local and meaningful only on the link where
227 Any DNS query for a name ending with ".local." MUST be sent
228 to the mDNS multicast address (224.0.0.251 or its IPv6 equivalent
232 Expires 10th February 2007 Cheshire & Krochmal [Page 4]
234 Internet Draft Multicast DNS 10th August 2006
237 It is unimportant whether a name ending with ".local." occurred
238 because the user explicitly typed in a fully qualified domain name
239 ending in ".local.", or because the user entered an unqualified
240 domain name and the host software appended the suffix ".local."
241 because that suffix appears in the user's search list. The ".local."
242 suffix could appear in the search list because the user manually
243 configured it, or because it was received in a DHCP option, or via
244 any other valid mechanism for configuring the DNS search list. In
245 this respect the ".local." suffix is treated no differently to any
246 other search domain that might appear in the DNS search list.
248 DNS queries for names that do not end with ".local." MAY be sent to
249 the mDNS multicast address, if no other conventional DNS server is
250 available. This can allow hosts on the same link to continue
251 communicating using each other's globally unique DNS names during
252 network outages which disrupt communication with the greater
253 Internet. When resolving global names via local multicast, it is even
254 more important to use DNSSEC or other security mechanisms to ensure
255 that the response is trustworthy. Resolving global names via local
256 multicast is a contentious issue, and this document does not discuss
257 it in detail, instead concentrating on the issue of resolving local
258 names using DNS packets sent to a multicast address.
260 A host that belongs to an organization or individual who has control
261 over some portion of the DNS namespace can be assigned a globally
262 unique name within that portion of the DNS namespace, for example,
263 "cheshire.apple.com." For those of us who have this luxury, this
264 works very well. However, the majority of home customers do not have
265 easy access to any portion of the global DNS namespace within which
266 they have the authority to create names as they wish. This leaves the
267 majority of home computers effectively anonymous for practical
270 To remedy this problem, this document allows any computer user to
271 elect to give their computers link-local Multicast DNS host names of
272 the form: "single-dns-label.local." For example, a laptop computer
273 may answer to the name "cheshire.local." Any computer user is granted
274 the authority to name their computer this way, provided that the
275 chosen host name is not already in use on that link. Having named
276 their computer this way, the user has the authority to continue using
277 that name until such time as a name conflict occurs on the link which
278 is not resolved in the user's favour. If this happens, the computer
279 (or its human user) SHOULD cease using the name, and may choose to
280 attempt to allocate a new unique name for use on that link. These
281 conflicts are expected to be relatively rare for people who choose
282 reasonably imaginative names, but it is still important to have a
283 mechanism in place to handle them when they happen.
285 The point made in the previous paragraph is very important and bears
286 repeating. It is easy for those of us in the IETF community who run
287 our own name servers at home to forget that the majority of computer
290 Expires 10th February 2007 Cheshire & Krochmal [Page 5]
292 Internet Draft Multicast DNS 10th August 2006
295 users do not run their own name server and have no easy way to create
296 their own host names. When these users wish to transfer files between
297 two laptop computers, they are frequently reduced to typing in
298 dotted-decimal IP addresses because they simply have no other way for
299 one host to refer to the other by name. This is a sorry state of
300 affairs. What is worse, most users don't even bother trying to use
301 dotted-decimal IP addresses. Most users still move data between
302 machines by burning it onto CD-R, copying it onto a USB "keychain"
303 flash drive, or similar removable media.
305 In a world of gigabit Ethernet and ubiquitous wireless networking it
306 is a sad indictment of the networking community that most users still
309 Allowing ad-hoc allocation of single-label names in a single flat
310 ".local." namespace may seem to invite chaos. However, operational
311 experience with AppleTalk NBP names [NBP], which on any given link
312 are also effectively single-label names in a flat namespace, shows
313 that in practice name collisions happen extremely rarely and are not
314 a problem. Groups of computer users from disparate organizations
315 bring Macintosh laptop computers to events such as IETF Meetings, the
316 Mac Hack conference, the Apple World Wide Developer Conference, etc.,
317 and complaints at these events about users suffering conflicts and
318 being forced to rename their machines have never been an issue.
320 This document advocates a single flat namespace for dot-local host
321 names, (i.e. the names of DNS address records), but other DNS record
322 types (such as those used by DNS Service Discovery [DNS-SD]) may
323 contain as many labels as appropriate for the desired usage, subject
324 to the 255-byte name length limit specified below in Section 3.3
325 "Maximum Multicast DNS Name Length".
327 Enforcing uniqueness of host names (i.e. the names of DNS address
328 records mapping names to IP addresses) is probably desirable in the
329 common case, but this document does not mandate that. It is
330 permissible for a collection of coordinated hosts to agree to
331 maintain multiple DNS address records with the same name, possibly
332 for load balancing or fault-tolerance reasons. This document does not
333 take a position on whether that is sensible. It is important that
334 both modes of operation are supported. The Multicast DNS protocol
335 allows hosts to verify and maintain unique names for resource records
336 where that behavior is desired, and it also allows hosts to maintain
337 multiple resource records with a single shared name where that
338 behavior is desired. This consideration applies to all resource
339 records, not just address records (host names). In summary: It is
340 required that the protocol have the ability to detect and handle name
341 conflicts, but it is not required that this ability be used for every
348 Expires 10th February 2007 Cheshire & Krochmal [Page 6]
350 Internet Draft Multicast DNS 10th August 2006
353 3.1 Governing Standards Body
355 Note that this use of the ".local." suffix falls under IETF/IANA
356 jurisdiction, not ICANN jurisdiction. DNS is an IETF network
357 protocol, governed by protocol rules defined by the IETF. These IETF
358 protocol rules dictate character set, maximum name length, packet
359 format, etc. ICANN determines additional rules that apply when the
360 IETF's DNS protocol is used on the public Internet. In contrast,
361 private uses of the DNS protocol on isolated private networks are not
362 governed by ICANN. Since this proposed change is a change to the core
363 DNS protocol rules, it affects everyone, not just those machines
364 using the ICANN-governed Internet. Hence this change falls into the
365 category of an IETF protocol rule, not an ICANN usage rule.
367 This allocation of responsibility is formally established in
368 "Memorandum of Understanding Concerning the Technical Work of the
369 Internet Assigned Numbers Authority" [RFC 2860]. Exception (a) of
370 clause 4.3 states that the IETF has the authority to instruct IANA
371 to reserve pseudo-TLDs as required for protocol design purposes.
372 For example, "Reserved Top Level DNS Names" [RFC 2606] defines
373 the following pseudo-TLDs:
381 3.2 Private DNS Namespaces
383 Note also that the special treatment of names ending in ".local." has
384 been implemented in Macintosh computers since the days of Mac OS 9,
385 and continues today in Mac OS X. There are also implementations for
386 Linux and other platforms [dotlocal]. Operators setting up private
387 internal networks ("intranets") are advised that their lives may be
388 easier if they avoid using the suffix ".local." in names in their
389 private internal DNS server. Alternative possibilities include:
398 Another alternative naming scheme, advocated by Professor D. J.
399 Bernstein, is to use a numerical suffix, such as ".6." [djbdl].
406 Expires 10th February 2007 Cheshire & Krochmal [Page 7]
408 Internet Draft Multicast DNS 10th August 2006
411 3.3 Maximum Multicast DNS Name Length
415 "the total number of octets that represent a domain name (i.e.,
416 the sum of all label octets and label lengths) is limited to 255."
418 This text implies that the final root label at the end of every name
419 is included in this count (a name can't be represented without it),
420 but the text does not explicitly state that. Implementations of
421 Multicast DNS MUST include the label length byte of the final root
422 label at the end of every name when enforcing the rule that no name
423 may be longer than 255 bytes. For example, the length of the name
424 "apple.com." is considered to be 11, which is the number of bytes it
425 takes to represent that name in a packet without using name
428 ------------------------------------------------------
429 | 0x05 | a | p | p | l | e | 0x03 | c | o | m | 0x00 |
430 ------------------------------------------------------
433 4. Source Address Check
435 All Multicast DNS responses (including responses sent via unicast)
436 SHOULD be sent with IP TTL set to 255. This is recommended to provide
437 backwards-compatibility with older Multicast DNS clients that check
438 the IP TTL on reception to determine whether the packet originated
439 on the local link. These older clients discard all packets with TTLs
442 A host sending Multicast DNS queries to a link-local destination
443 address (including the 224.0.0.251 link-local multicast address)
444 MUST only accept responses to that query that originate from the
445 local link, and silently discard any other response packets. Without
446 this check, it could be possible for remote rogue hosts to send
447 spoof answer packets (perhaps unicast to the victim host) which the
448 receiving machine could misinterpret as having originated on the
451 The test for whether a response originated on the local link
454 * All responses sent to the link-local multicast address 224.0.0.251
455 are necessarily deemed to have originated on the local link,
456 regardless of source IP address. This is essential to allow devices
457 to work correctly and reliably in unusual configurations, such as
458 multiple logical IP subnets overlayed on a single link, or in cases
459 of severe misconfiguration, where devices are physically connected
460 to the same link, but are currently misconfigured with completely
461 unrelated IP addresses and subnet masks.
464 Expires 10th February 2007 Cheshire & Krochmal [Page 8]
466 Internet Draft Multicast DNS 10th August 2006
469 * For responses sent to a unicast destination address, the source IP
470 address in the packet is checked to see if it is an address on a
471 local subnet. An address is determined to be on a local subnet if,
472 for (one of) the address(es) configured on the interface receiving
473 the packet, (I & M) == (P & M), where I and M are the interface
474 address and subnet mask respectively, P is the source IP address
475 from the packet, '&' represents the bitwise logical 'and'
476 operation, and '==' represents a bitwise equality test.
478 Since queriers will ignore responses apparently originating outside
479 the local subnet, responders SHOULD avoid generating responses that
480 it can reasonably predict will be ignored. This applies particularly
481 in the case of overlayed subnets. If a responder receives a query
482 addressed to the link-local multicast address 224.0.0.251, from a
483 source address not apparently on the same subnet as the responder,
484 then even if the query indicates that a unicast response is preferred
485 (see Section 6.5, "Questions Requesting Unicast Responses"), the
486 responder SHOULD elect to respond by multicast anyway, since it can
487 reasonably predict that a unicast response with an apparently
488 non-local source address will probably be ignored.
491 5. Reverse Address Mapping
493 Like ".local.", the IPv4 and IPv6 reverse mapping domains are also
494 defined to be link-local.
496 Any DNS query for a name ending with "254.169.in-addr.arpa." MUST
497 be sent to the mDNS multicast address 224.0.0.251. Since names under
498 this domain correspond to IPv4 link-local addresses, it is logical
499 that the local link is the best place to find information pertaining
502 Likewise, any DNS query for a name within the reverse mapping domains
503 for IPv6 Link-Local addresses ("8.e.f.ip6.arpa.", "9.e.f.ip6.arpa.",
504 "a.e.f.ip6.arpa.", and "b.e.f.ip6.arpa.") MUST be sent to the IPv6
505 mDNS link-local multicast address FF02::FB.
522 Expires 10th February 2007 Cheshire & Krochmal [Page 9]
524 Internet Draft Multicast DNS 10th August 2006
529 There are three kinds of Multicast DNS Queries, one-shot queries of
530 the kind made by today's conventional DNS clients, one-shot queries
531 accumulating multiple responses made by multicast-aware DNS clients,
532 and continuous ongoing Multicast DNS Queries used by IP network
535 A Multicast DNS Responder that is offering records that are intended
536 to be unique on the local link MUST also implement a Multicast DNS
537 Querier so that it can first verify the uniqueness of those records
538 before it begins answering queries for them.
541 6.1 One-Shot Multicast DNS Queries
543 An unsophisticated DNS client may simply send its DNS queries blindly
544 to 224.0.0.251:5353, without necessarily even being aware what a
545 multicast address is. This change can typically be implemented with
546 just a few lines of code in an existing DNS resolver library. Any
547 time the name being queried for falls within one of the reserved
548 mDNS domains (see Section 12 "Special Characteristics of Multicast
549 DNS Domains") the query is sent to 224.0.0.251:5353 instead of the
550 configured unicast DNS server address that would otherwise be used.
551 Typically the timeout would also be shortened to two or three
552 seconds, but it's possible to make a minimal mDNS client with no
553 other changes apart from these.
555 A simple DNS client like this will typically just take the first
556 response it receives. It will not listen for additional UDP
557 responses, but in many instances this may not be a serious problem.
558 If a user types "http://cheshire.local." into their Web browser and
559 gets to see the page they were hoping for, then the protocol has met
560 the user's needs in this case.
562 While an unsophisticated DNS client like this is perfectly adequate
563 for simple hostname lookup, it may not get ideal behavior in
564 other cases. Additional refinements that may be adopted by more
565 sophisticated clients are described below.
568 6.2 One-Shot Queries, Accumulating Multiple Responses
570 A more sophisticated DNS client should understand that Multicast DNS
571 is not exactly the same as unicast DNS, and should modify its
572 behavior in some simple ways.
574 As described above, there are some cases, such as looking up the
575 address associated with a unique host name, where a single response
576 is sufficient, and moreover may be all that is expected. However,
577 there are other DNS queries where more than one response is
580 Expires 10th February 2007 Cheshire & Krochmal [Page 10]
582 Internet Draft Multicast DNS 10th August 2006
585 possible, and for these queries a more sophisticated Multicast DNS
586 client should include the ability to wait for an appropriate period
587 of time to collect multiple responses.
589 A naive DNS client retransmits its query only so long as it has
590 received no response. A more sophisticated Multicast DNS client is
591 aware that having received one response is not necessarily an
592 indication that it might not receive others, and has the ability to
593 retransmit its query an appropriate number of times at appropriate
594 intervals until it is satisfied with the collection of responses it
597 A more sophisticated Multicast DNS client that is retransmitting
598 a query for which it has already received some responses, MUST
599 implement Known Answer Suppression, as described below in Section 7.1
600 "Known Answer Suppression". This indicates to responders who have
601 already replied that their responses have been received, and they
602 don't need to send them again in response to this repeated query. In
603 addition, when retransmitting queries, the interval between the first
604 two queries SHOULD be one second, and the intervals between
605 subsequent queries SHOULD double.
608 6.3 Continuous Multicast DNS Querying
610 In One-Shot Queries, with either a single or multiple responses,
611 the underlying assumption is that the transaction begins when the
612 application issues a query, and ends when all the desired responses
613 have been received. There is another type of operation which is more
614 akin to continuous monitoring.
616 iTunes users are accustomed to seeing a list of shared network music
617 libraries in the sidebar of the iTunes window. There is no "refresh"
618 button for the user to click because the list is always accurate,
619 always reflecting the currently available libraries. When a new
620 library becomes available it promptly appears in the list, and when
621 a library becomes unavailable it promptly disappears. It is vitally
622 important that this responsive user interface be achieved without
623 naive polling that would place an unreasonable burden on the network.
625 Therefore, when retransmitting mDNS queries to implement this kind
626 of continuous monitoring, the interval between the first two queries
627 SHOULD be one second, the intervals between the subsequent queries
628 SHOULD double, and the querier MUST implement Known Answer
629 Suppression, as described below in Section 7.1. When the interval
630 between queries reaches or exceeds 60 minutes, a querier MAY cap the
631 interval to a maximum of 60 minutes, and perform subsequent queries
632 at a steady-state rate of one query per hour. To avoid accidental
633 synchronization when for some reason multiple clients begin querying
634 at exactly the same moment (e.g. because of some common external
635 trigger event), a Multicast DNS Querier SHOULD also delay the first
638 Expires 10th February 2007 Cheshire & Krochmal [Page 11]
640 Internet Draft Multicast DNS 10th August 2006
643 query of the series by a randomly-chosen amount in the range
646 When a Multicast DNS Querier receives an answer, the answer contains
647 a TTL value that indicates for how many seconds this answer is valid.
648 After this interval has passed, the answer will no longer be valid
649 and SHOULD be deleted from the cache. Before this time is reached,
650 a Multicast DNS Querier which has clients with an active interest in
651 the state of that record (e.g. a network browsing window displaying
652 a list of discovered services to the user) SHOULD re-issue its query
653 to determine whether the record is still valid.
655 To perform this cache maintenance, a Multicast DNS Querier should
656 plan to re-query for records after at least 50% of the record
657 lifetime has elapsed. This document recommends the following
660 The Querier should plan to issue a query at 80% of the record
661 lifetime, and then if no answer is received, at 85%, 90% and 95%.
662 If an answer is received, then the remaining TTL is reset to the
663 value given in the answer, and this process repeats for as long as
664 the Multicast DNS Querier has an ongoing interest in the record.
665 If after four queries no answer is received, the record is deleted
666 when it reaches 100% of its lifetime. A Multicast DNS Querier MUST
667 NOT perform this cache maintenance for records for which it has no
668 clients with an active interest. If the expiry of a particular record
669 from the cache would result in no net effect to any client software
670 running on the Querier device, and no visible effect to the human
671 user, then there is no reason for the Multicast DNS Querier to
672 waste network bandwidth checking whether the record remains valid.
674 To avoid the case where multiple Multicast DNS Queriers on a network
675 all issue their queries simultaneously, a random variation of 2% of
676 the record TTL should be added, so that queries are scheduled to be
677 performed at 80-82%, 85-87%, 90-92% and then 95-97% of the TTL.
680 6.4 Multiple Questions per Query
682 Multicast DNS allows a querier to place multiple questions in the
683 Question Section of a single Multicast DNS query packet.
685 The semantics of a Multicast DNS query packet containing multiple
686 questions is identical to a series of individual DNS query packets
687 containing one question each. Combining multiple questions into a
688 single packet is purely an efficiency optimization, and has no other
689 semantic significance.
696 Expires 10th February 2007 Cheshire & Krochmal [Page 12]
698 Internet Draft Multicast DNS 10th August 2006
701 6.5 Questions Requesting Unicast Responses
703 Sending Multicast DNS responses via multicast has the benefit that
704 all the other hosts on the network get to see those responses, and
705 can keep their caches up to date, and detect conflicting responses.
707 However, there are situations where all the other hosts on the
708 network don't need to see every response. Some examples are a laptop
709 computer waking from sleep, or the Ethernet cable being connected to
710 a running machine, or a previously inactive interface being activated
711 through a configuration change. At the instant of wake-up or link
712 activation, the machine is a brand new participant on a new network.
713 Its Multicast DNS cache for that interface is empty, and it has no
714 knowledge of its peers on that link. It may have a significant number
715 of questions that it wants answered right away to discover
716 information about its new surroundings and present that information
717 to the user. As a new participant on the network, it has no idea
718 whether the exact same questions may have been asked and answered
719 just seconds ago. In this case, triggering a large sudden flood of
720 multicast responses may impose an unreasonable burden on the network.
722 To avoid large floods of potentially unnecessary responses in these
723 cases, Multicast DNS defines the top bit in the class field of a DNS
724 question as the "unicast response" bit. When this bit is set in a
725 question, it indicates that the Querier is willing to accept unicast
726 responses instead of the usual multicast responses. These questions
727 requesting unicast responses are referred to as "QU" questions, to
728 distinguish them from the more usual questions requesting multicast
729 responses ("QM" questions). A Multicast DNS Querier sending its
730 initial batch of questions immediately on wake from sleep or
731 interface activation SHOULD set the "QU" bit in those questions.
733 When a question is retransmitted (as described in Section 6.3
734 "Continuous Multicast DNS Querying") the "QU" bit SHOULD NOT be set
735 in subsequent retransmissions of that question. Subsequent
736 retransmissions SHOULD be usual "QM" questions. After the first
737 question has received its responses, the querier should have a large
738 known-answer list (see "Known Answer Suppression" below) so that
739 subsequent queries should elicit few, if any, further responses.
740 Reverting to multicast responses as soon as possible is important
741 because of the benefits that multicast responses provide (see
742 "Benefits of Multicast Responses" below). In addition, the "QU" bit
743 SHOULD be set only for questions that are active and ready to be sent
744 the moment of wake from sleep or interface activation. New questions
745 issued by clients afterwards should be treated as normal "QM"
746 questions and SHOULD NOT have the "QU" bit set on the first question
749 When receiving a question with the "unicast response" bit set, a
750 responder SHOULD usually respond with a unicast packet directed back
751 to the querier. If the responder has not multicast that record
754 Expires 10th February 2007 Cheshire & Krochmal [Page 13]
756 Internet Draft Multicast DNS 10th August 2006
759 recently (within one quarter of its TTL), then the responder SHOULD
760 instead multicast the response so as to keep all the peer caches up
761 to date, and to permit passive conflict detection. In the case of
762 answering a probe question with the "unicast response" bit set, the
763 responder should always generate the requested unicast response, but
764 may also send a multicast announcement too if the time since the last
765 multicast announcement of that record is more than a quarter of its
768 Except when defending a unique name against a probe from another host
769 unicast replies are subject to all the same packet generation rules
770 as multicast replies, including the cache flush bit (see Section
771 11.3, "Announcements to Flush Outdated Cache Entries") and randomized
772 delays to reduce network collisions (see Section 8, "Responding").
775 6.6 Delaying Initial Query
777 If a query is issued for which there already exist one or more
778 records in the local cache, and those record(s) were received with
779 the cache flush bit set (see Section 11.3, "Announcements to Flush
780 Outdated Cache Entries"), indicating that they form a unique RRSet,
781 then the host SHOULD delay its initial query by imposing a random
782 delay from 500-1000ms. This is to avoid the situation where a group
783 of hosts are synchronized by some external event and all perform
784 the same query simultaneously. This means that when the first host
785 (selected randomly by this algorithm) transmits its query, all the
786 other hosts that were about to transmit the same query can suppress
787 their superfluous queries, as described in "Duplicate Question
791 6.7 Direct Unicast Queries to port 5353
793 In specialized applications there may be rare situations where it
794 makes sense for a Multicast DNS Querier to send its query via unicast
795 to a specific machine. When a Multicast DNS Responder receives a
796 query via direct unicast, it SHOULD respond as it would for a
797 "QU" query, as described above in Section 6.5 "Questions Requesting
798 Unicast Responses". Since it is possible for a unicast query to be
799 received from a machine outside the local link, Responders SHOULD
800 check that the source address in the query packet matches the local
801 subnet for that link, and silently ignore the packet if not.
803 There may be specialized situations, outside the scope of this
804 document, where it is intended and desirable to create a Responder
805 that does answer queries originating outside the local link. Such
806 a Responder would need to ensure that these non-local queries are
807 always answered via unicast back to the Querier, since an answer sent
808 via link-local multicast would not reach a Querier outside the local
812 Expires 10th February 2007 Cheshire & Krochmal [Page 14]
814 Internet Draft Multicast DNS 10th August 2006
817 7. Duplicate Suppression
819 A variety of techniques are used to reduce the amount of redundant
820 traffic on the network.
822 7.1 Known Answer Suppression
824 When a Multicast DNS Querier sends a query to which it already knows
825 some answers, it populates the Answer Section of the DNS message with
828 A Multicast DNS Responder SHOULD NOT answer a Multicast DNS Query if
829 the answer it would give is already included in the Answer Section
830 with an RR TTL at least half the correct value. If the RR TTL of the
831 answer as given in the Answer Section is less than half of the true
832 RR TTL as known by the Multicast DNS Responder, the responder MUST
833 send an answer so as to update the Querier's cache before the record
834 becomes in danger of expiration.
836 Because a Multicast DNS Responder will respond if the remaining TTL
837 given in the known answer list is less than half the true TTL, it is
838 superfluous for the Querier to include such records in the known
839 answer list. Therefore a Multicast DNS Querier SHOULD NOT include
840 records in the known answer list whose remaining TTL is less than
841 half their original TTL. Doing so would simply consume space in the
842 packet without achieving the goal of suppressing responses, and would
843 therefore be a pointless waste of network bandwidth.
845 A Multicast DNS Querier MUST NOT cache resource records observed in
846 the Known Answer Section of other Multicast DNS Queries. The Answer
847 Section of Multicast DNS Queries is not authoritative. By placing
848 information in the Answer Section of a Multicast DNS Query the
849 querier is stating that it *believes* the information to be true.
850 It is not asserting that the information *is* true. Some of those
851 records may have come from other hosts that are no longer on the
852 network. Propagating that stale information to other Multicast DNS
853 Queriers on the network would not be helpful.
856 7.2 Multi-Packet Known Answer Suppression
858 Sometimes a Multicast DNS Querier will already have too many answers
859 to fit in the Known Answer Section of its query packets. In this
860 case, it should issue a Multicast DNS Query containing a question and
861 as many Known Answer records as will fit. It MUST then set the TC
862 (Truncated) bit in the header before sending the Query. It MUST then
863 immediately follow the packet with another query packet containing no
864 questions, and as many more Known Answer records as will fit. If
865 there are still too many records remaining to fit in the packet, it
866 again sets the TC bit and continues until all the Known Answer
867 records have been sent.
870 Expires 10th February 2007 Cheshire & Krochmal [Page 15]
872 Internet Draft Multicast DNS 10th August 2006
875 A Multicast DNS Responder seeing a Multicast DNS Query with the TC
876 bit set defers its response for a time period randomly selected in
877 the interval 400-500ms. This gives the Multicast DNS Querier time to
878 send additional Known Answer packets before the Responder responds.
879 If the Responder sees any of its answers listed in the Known Answer
880 lists of subsequent packets from the querying host, it SHOULD delete
881 that answer from the list of answers it is planning to give, provided
882 that no other host on the network is also waiting to receive the same
885 If the Responder receives additional Known Answer packets with the TC
886 bit set, it SHOULD extend the delay as necessary to ensure a pause of
887 400-500ms after the last such packet before it sends its answer. This
888 opens the potential risk that a continuous stream of Known Answer
889 packets could, theoretically, prevent a Responder from answering
890 indefinitely. In practice answers are never actually delayed
891 significantly, and should a situation arise where significant delays
892 did happen, that would be a scenario where the network is so
893 overloaded that it would be desirable to err on the side of caution.
894 The consequence of delaying an answer may be that it takes a user
895 longer than usual to discover all the services on the local network;
896 in contrast the consequence of incorrectly answering before all the
897 Known Answer packets have been received would be wasting bandwidth
898 sending unnecessary answers on an already overloaded network. In this
899 (rare) situation, sacrificing speed to preserve reliable network
900 operation is the right trade-off.
903 7.3 Duplicate Question Suppression
905 If a host is planning to send a query, and it sees another host on
906 the network send a query containing the same question, and the Known
907 Answer Section of that query does not contain any records which this
908 host would not also put in its own Known Answer Section, then this
909 host should treat its own query as having been sent. When multiple
910 clients on the network are querying for the same resource records,
911 there is no need for them to all be repeatedly asking the same
915 7.4 Duplicate Answer Suppression
917 If a host is planning to send an answer, and it sees another host on
918 the network send a response packet containing the same answer record,
919 and the TTL in that record is not less than the TTL this host would
920 have given, then this host should treat its own answer as having been
921 sent. When multiple responders on the network have the same data,
922 there is no need for all of them to respond.
924 This feature is particularly useful when multiple Sleep Proxy Servers
925 are deployed (see Section 17, "Multicast DNS and Power Management").
928 Expires 10th February 2007 Cheshire & Krochmal [Page 16]
930 Internet Draft Multicast DNS 10th August 2006
933 In the future it is possible that every general-purpose OS (Mac,
934 Windows, Linux, etc.) will implement Sleep Proxy Service as a matter
935 of course. In this case there could be a large number of Sleep Proxy
936 Servers on any given network, which is good for reliability and
937 fault-tolerance, but would be bad for the network if every Sleep
938 Proxy Server were to answer every query.
942 When a Multicast DNS Responder constructs and sends a Multicast DNS
943 response packet, the Answer Section of that packet must contain only
944 records for which that Responder is explicitly authoritative. These
945 answers may be generated because the record answers a question
946 received in a Multicast DNS query packet, or at certain other times
947 that the responder determines than an unsolicited announcement is
948 warranted. A Multicast DNS Responder MUST NOT place records from its
949 cache, which have been learned from other responders on the network,
950 in the Answer Section of outgoing response packets. Only an
951 authoritative source for a given record is allowed to issue responses
952 containing that record.
954 The determination of whether a given record answers a given question
955 is done using the standard DNS rules: The record name must match the
956 question name, the record rrtype must match the question qtype
957 (unless the qtype is "ANY"), and the record rrclass must match the
958 question qclass (unless the qclass is "ANY").
960 A Multicast DNS Responder MUST only respond when it has a positive
961 non-null response to send. Error responses must never be sent. The
962 non-existence of any name in a Multicast DNS Domain is ascertained by
963 the failure of any machine to respond to the Multicast DNS query, not
966 Multicast DNS Responses MUST NOT contain any questions in the
967 Question Section. Any questions in the Question Section of a received
968 Multicast DNS Response MUST be silently ignored. Multicast DNS
969 Queriers receiving Multicast DNS Responses do not care what question
970 elicited the response; they care only that the information in the
971 response is true and accurate.
973 A Multicast DNS Responder on Ethernet [IEEE802] and similar shared
974 multiple access networks SHOULD have the capability of delaying its
975 responses by up to 500ms, as determined by the rules described below.
976 If a large number of Multicast DNS Responders were all to respond
977 immediately to a particular query, a collision would be virtually
978 guaranteed. By imposing a small random delay, the number of
979 collisions is dramatically reduced. On a full-sized Ethernet using
980 the maximum cable lengths allowed and the maximum number of repeaters
981 allowed, an Ethernet frame is vulnerable to collisions during the
982 transmission of its first 256 bits. On 10Mb/s Ethernet, this equates
983 to a vulnerable time window of 25.6us. On higher-speed variants of
984 Ethernet, the vulnerable time window is shorter.
986 Expires 10th February 2007 Cheshire & Krochmal [Page 17]
988 Internet Draft Multicast DNS 10th August 2006
991 In the case where a Multicast DNS Responder has good reason to
992 believe that it will be the only responder on the link with a
993 positive non-null response (i.e. because it is able to answer every
994 question in the query packet, and for all of those answer records it
995 has previously verified that the name, rrtype and rrclass are unique
996 on the link) it SHOULD NOT impose any random delay before responding,
997 and SHOULD normally generate its response within at most 10ms.
998 In particular, this applies to responding to probe queries with the
999 "unicast response" bit set. Since receiving a probe query gives a
1000 clear indication that some other Responder is planning to start using
1001 this name in the very near future, answering such probe queries
1002 to defend a unique record is a high priority and needs to be done
1003 immediately, without delay. A probe query can be distinguished from
1004 a normal query by the fact that a probe query contains a proposed
1005 record in the Authority Section which answers the question in the
1006 Question Section (for more details, see Section 9.1, "Probing").
1008 Responding immediately without delay is appropriate for records like
1009 the address record for a particular host name, when the host name has
1010 been previously verified unique. Responding immediately without delay
1011 is *not* appropriate for things like looking up PTR records used for
1012 DNS Service Discovery [DNS-SD], where a large number of responses may
1015 In any case where there may be multiple responses, such as queries
1016 where the answer is a member of a shared resource record set, each
1017 responder SHOULD delay its response by a random amount of time
1018 selected with uniform random distribution in the range 20-120ms.
1019 The reason for requiring that the delay be at least 20ms is to
1020 accommodate the situation where two or more query packets are sent
1021 back-to-back, because in that case we want a Responder with answers
1022 to more than one of those queries to have the opportunity to
1023 aggregate all of its answers into a single response packet.
1025 In the case where the query has the TC (truncated) bit set,
1026 indicating that subsequent known answer packets will follow,
1027 responders SHOULD delay their responses by a random amount of time
1028 selected with uniform random distribution in the range 400-500ms,
1029 to allow enough time for all the known answer packets to arrive,
1030 as described in Section 7.2 "Multi-Packet Known Answer Suppression".
1032 Except when a unicast response has been explicitly requested via the
1033 "unicast response" bit, Multicast DNS Responses MUST be sent to UDP
1034 port 5353 (the well-known port assigned to mDNS) on the 224.0.0.251
1035 multicast address (or its IPv6 equivalent FF02::FB). Operating in a
1036 Zeroconf environment requires constant vigilance. Just because a name
1037 has been previously verified unique does not mean it will continue
1038 to be so indefinitely. By allowing all Multicast DNS Responders to
1039 constantly monitor their peers' responses, conflicts arising out
1040 of network topology changes can be promptly detected and resolved.
1044 Expires 10th February 2007 Cheshire & Krochmal [Page 18]
1046 Internet Draft Multicast DNS 10th August 2006
1049 Sending all responses by multicast also facilitates opportunistic
1050 caching by other hosts on the network.
1052 To protect the network against excessive packet flooding due to
1053 software bugs or malicious attack, a Multicast DNS Responder MUST NOT
1054 (except in the one special case of answering probe queries) multicast
1055 a record on a given interface until at least one second has elapsed
1056 since the last time that record was multicast on that particular
1057 interface. A legitimate client on the network should have seen the
1058 previous transmission and cached it. A client that did not receive
1059 and cache the previous transmission will retry its request and
1060 receive a subsequent response. In the special case of answering probe
1061 queries, because of the limited time before the probing host will
1062 make its decision about whether or not to use the name, a Multicast
1063 DNS Responder MUST respond quickly. In this special case only, when
1064 responding via multicast to a probe, a Multicast DNS Responder is
1065 only required to delay its transmission as necessary to ensure an
1066 interval of at least 250ms since the last time the record was
1067 multicast on that interface.
1070 8.2 Multi-Question Queries
1072 Multicast DNS Responders MUST correctly handle DNS query packets
1073 containing more than one question, by answering any or all of the
1074 questions to which they have answers. Any (non-defensive) answers
1075 generated in response to query packets containing more than one
1076 question SHOULD be randomly delayed in the range 20-120ms, or
1077 400-500ms if the TC (truncated) bit is set, as described above.
1078 (Answers defending a name, in response to a probe for that name,
1079 are not subject to this delay rule and are still sent immediately.)
1082 8.2 Response Aggregation
1084 When possible, a responder SHOULD, for the sake of network
1085 efficiency, aggregate as many responses as possible into a single
1086 Multicast DNS response packet. For example, when a responder has
1087 several responses it plans to send, each delayed by a different
1088 interval, then earlier responses SHOULD be delayed by up to an
1089 additional 500ms if that will permit them to be aggregated with
1090 other responses scheduled to go out a little later.
1102 Expires 10th February 2007 Cheshire & Krochmal [Page 19]
1104 Internet Draft Multicast DNS 10th August 2006
1107 8.3 Legacy Unicast Responses
1109 If the source UDP port in a received Multicast DNS Query is not port
1110 5353, this indicates that the client originating the query is a
1111 simple client that does not fully implement all of Multicast DNS.
1112 In this case, the Multicast DNS Responder MUST send a UDP response
1113 directly back to the client, via unicast, to the query packet's
1114 source IP address and port. This unicast response MUST be a
1115 conventional unicast response as would be generated by a conventional
1116 unicast DNS server; for example, it MUST repeat the query ID and the
1117 question given in the query packet.
1119 The resource record TTL given in a legacy unicast response SHOULD NOT
1120 be greater than ten seconds, even if the true TTL of the Multicast
1121 DNS resource record is higher. This is because Multicast DNS
1122 Responders that fully participate in the protocol use the cache
1123 coherency mechanisms described in Section 11 "Resource Record TTL
1124 Values and Cache Coherency" to update and invalidate stale data. Were
1125 unicast responses sent to legacy clients to use the same high TTLs,
1126 these legacy clients, which do not implement these cache coherency
1127 mechanisms, could retain stale cached resource record data long after
1128 it is no longer valid.
1130 Having sent this unicast response, if the Responder has not sent this
1131 record in any multicast response recently, it SHOULD schedule the
1132 record to be sent via multicast as well, to facilitate passive
1133 conflict detection. "Recently" in this context means "if the time
1134 since the record was last sent via multicast is less than one quarter
1135 of the record's TTL".
1137 Note that while legacy queries usually contain exactly one question,
1138 they are permitted to contain multiple questions, and responders
1139 listening for multicast queries on 224.0.0.251:5353 MUST be prepared
1140 to handle this correctly, responding by generating a unicast response
1141 containing the list of question(s) they are answering in the Question
1142 Section, and the records answering those question(s) in the Answer
1146 9. Probing and Announcing on Startup
1148 Typically a Multicast DNS Responder should have, at the very least,
1149 address records for all of its active interfaces. Creating and
1150 advertising an HINFO record on each interface as well can be useful
1151 to network administrators.
1153 Whenever a Multicast DNS Responder starts up, wakes up from sleep,
1154 receives an indication of an Ethernet "Link Change" event, or has any
1155 other reason to believe that its network connectivity may have
1156 changed in some relevant way, it MUST perform the two startup steps
1160 Expires 10th February 2007 Cheshire & Krochmal [Page 20]
1162 Internet Draft Multicast DNS 10th August 2006
1167 The first startup step is that for all those resource records that a
1168 Multicast DNS Responder desires to be unique on the local link, it
1169 MUST send a Multicast DNS Query asking for those resource records, to
1170 see if any of them are already in use. The primary example of this is
1171 its address record which maps its unique host name to its unique IP
1172 address. All Probe Queries SHOULD be done using the desired resource
1173 record name and query type T_ANY (255), to elicit answers for all
1174 types of records with that name. This allows a single question to be
1175 used in place of several questions, which is more efficient on the
1176 network. It also allows a host to verify exclusive ownership of a
1177 name, which is desirable in most cases. It would be confusing, for
1178 example, if one host owned the "A" record for "myhost.local.", but
1179 a different host owned the HINFO record for that name.
1181 The ability to place more than one question in a Multicast DNS Query
1182 is useful here, because it can allow a host to use a single packet
1183 for all of its resource records instead of needing a separate packet
1184 for each. For example, a host can simultaneously probe for uniqueness
1185 of its "A" record and all its SRV records [DNS-SD] in the same query
1188 When ready to send its mDNS probe packet(s) the host should first
1189 wait for a short random delay time, uniformly distributed in the
1190 range 0-250ms. This random delay is to guard against the case where a
1191 group of devices are powered on simultaneously, or a group of devices
1192 are connected to an Ethernet hub which is then powered on, or some
1193 other external event happens that might cause a group of hosts to all
1194 send synchronized probes.
1196 250ms after the first query the host should send a second, then
1197 250ms after that a third. If, by 250ms after the third probe, no
1198 conflicting Multicast DNS responses have been received, the host may
1199 move to the next step, announcing. (Note that this is the one
1200 exception from the normal rule that there should be at least one
1201 second between repetitions of the same question, and the interval
1202 between subsequent repetitions should double.)
1204 When sending probe queries, a host MUST NOT consult its cache for
1205 potential answers. Only conflicting Multicast DNS responses received
1206 "live" from the network are considered valid for the purposes of
1207 determining whether probing has succeeded or failed.
1209 In order to allow services to announce their presence without
1210 unreasonable delay, the time window for probing is intentionally set
1211 quite short. As a result of this, from the time the first probe
1212 packet is sent, another device on the network using that name has
1213 just 750ms to respond to defend its name. On networks that are slow,
1214 or busy, or both, it is possible for round-trip latency to account
1215 for a few hundred milliseconds, and software delays in slow devices
1218 Expires 10th February 2007 Cheshire & Krochmal [Page 21]
1220 Internet Draft Multicast DNS 10th August 2006
1223 can add additional delay. For this reason, it is important that when
1224 a device receives a probe query for a name that it is currently using
1225 for unique records, it SHOULD generate its response to defend that
1226 name immediately and send it as quickly as possible. The usual rules
1227 about random delays before responding, to avoid sudden bursts of
1228 simultaneous answers from different hosts, do not apply here since
1229 at most one host should ever respond to a given probe question. Even
1230 when a single DNS query packet contains multiple probe questions,
1231 it would be unusual for that packet to elicit a defensive response
1232 from more than one other host. Because of the mDNS multicast rate
1233 limiting rules, the first two probes SHOULD be sent as "QU" questions
1234 with the "unicast response" bit set, to allow a defending host to
1235 respond immediately via unicast, instead of potentially having to
1236 wait before replying via multicast. At the present time, this
1237 document recommends that the third probe SHOULD be sent as a standard
1238 "QM" question, for backwards compatibility with the small number of
1239 old devices still in use that don't implement unicast responses.
1241 If, at any time during probing, from the beginning of the initial
1242 random 0-250ms delay onward, any conflicting Multicast DNS responses
1243 are received, then the probing host MUST defer to the existing host,
1244 and MUST choose new names for some or all of its resource records
1245 as appropriate, to avoid conflict with pre-existing hosts on the
1246 network. In the case of a host probing using query type T_ANY as
1247 recommended above, any answer containing a record with that name,
1248 of any type, MUST be considered a conflicting response and handled
1251 If fifteen failures occur within any ten-second period, then the host
1252 MUST wait at least five seconds before each successive additional
1253 probe attempt. This is to help ensure that in the event of software
1254 bugs or other unanticipated problems, errant hosts do not flood the
1255 network with a continuous stream of multicast traffic. For very
1256 simple devices, a valid way to comply with this requirement is
1257 to always wait five seconds after any failed probe attempt before
1260 If a responder knows by other means, with absolute certainty, that
1261 its unique resource record set name, rrtype and rrclass cannot
1262 already be in use by any other responder on the network, then it
1263 MAY skip the probing step for that resource record set. For example,
1264 when creating the reverse address mapping PTR records, the host can
1265 reasonably assume that no other host will be trying to create those
1266 same PTR records, since that would imply that the two hosts were
1267 trying to use the same IP address, and if that were the case, the
1268 two hosts would be suffering communication problems beyond the scope
1269 of what Multicast DNS is designed to solve.
1276 Expires 10th February 2007 Cheshire & Krochmal [Page 22]
1278 Internet Draft Multicast DNS 10th August 2006
1281 9.2 Simultaneous Probe Tie-Breaking
1283 The astute reader will observe that there is a race condition
1284 inherent in the previous description. If two hosts are probing for
1285 the same name simultaneously, neither will receive any response to
1286 the probe, and the hosts could incorrectly conclude that they may
1287 both proceed to use the name. To break this symmetry, each host
1288 populates the Query packets's Authority Section with the record or
1289 records with the rdata that it would be proposing to use, should its
1290 probing be successful. The Authority Section is being used here in a
1291 way analogous to the way it is used as the "Update Section" in a DNS
1292 Update packet [RFC 2136].
1294 When a host is probing for a group of related records with the same
1295 name (e.g. the SRV and TXT record describing a DNS-SD service), only
1296 a single question need be placed in the Question Section, since query
1297 type T_ANY (255) is used, which will elicit answers for all records
1298 with that name. However, for tie-breaking to work correctly in all
1299 cases, the Authority Section must contain *all* the records and
1300 proposed rdata being probed for uniqueness.
1302 When a host that is probing for a record sees another host issue a
1303 query for the same record, it consults the Authority Section of that
1304 query. If it finds any resource record(s) there which answers the
1305 query, then it compares the data of that (those) resource record(s)
1306 with its own tentative data. We consider first the simple case of a
1307 host probing for a single record, receiving a simultaneous probe from
1308 another host also probing for a single record. The two records are
1309 compared and the lexicographically later data wins. This means that
1310 if the host finds that its own data is lexicographically later, it
1311 simply ignores the other host's probe. If the host finds that its own
1312 data is lexicographically earlier, then it treats this exactly as if
1313 it had received a positive answer to its query, and concludes that it
1314 may not use the desired name.
1316 The determination of "lexicographically later" is performed by first
1317 comparing the record class, then the record type, then raw comparison
1318 of the binary content of the rdata without regard for meaning or
1319 structure. If the record classes differ, then the numerically greater
1320 class is considered "lexicographically later". Otherwise, if the
1321 record types differ, then the numerically greater type is considered
1322 "lexicographically later". If the rrtype and rrclass both match then
1323 the rdata is compared.
1325 In the case of resource records containing rdata that is subject to
1326 name compression, the names MUST be uncompressed before comparison.
1327 (The details of how a particular name is compressed is an artifact of
1328 how and where the record is written into the DNS message; it is not
1329 an intrinsic property of the resource record itself.)
1334 Expires 10th February 2007 Cheshire & Krochmal [Page 23]
1336 Internet Draft Multicast DNS 10th August 2006
1339 The bytes of the raw uncompressed rdata are compared in turn,
1340 interpreting the bytes as eight-bit UNSIGNED values, until a byte
1341 is found whose value is greater than that of its counterpart (in
1342 which case the rdata whose byte has the greater value is deemed
1343 lexicographically later) or one of the resource records runs out
1344 of rdata (in which case the resource record which still has
1345 remaining data first is deemed lexicographically later).
1347 The following is an example of a conflict:
1349 cheshire.local. A 169.254.99.200
1350 cheshire.local. A 169.254.200.50
1352 In this case 169.254.200.50 is lexicographically later (the third
1353 byte, with value 200, is greater than its counterpart with value 99),
1354 so it is deemed the winner.
1356 Note that it is vital that the bytes are interpreted as UNSIGNED
1357 values in the range 0-255, or the wrong outcome may result. In
1358 the example above, if the byte with value 200 had been incorrectly
1359 interpreted as a signed eight-bit value then it would be interpreted
1360 as value -56, and the wrong address record would be deemed the
1364 9.2.1 Simultaneous Probe Tie-Breaking for Multiple Records
1366 When a host is probing for a set of records with the same name, or a
1367 packet is received containing multiple tie-breaker records answering
1368 a given probe question in the Question Section, the host's records
1369 and the tie-breaker records from the packet are each sorted into
1370 order, and then compared pairwise, using the same comparison
1371 technique described above, until a difference is found.
1373 The records are sorted using the same lexicographical order as
1374 described above, that is: if the record classes differ, the record
1375 with the lower class number comes first. If the classes are the same
1376 but the rrtypes differ, the record with the lower rrtype number comes
1377 first. If the class and rrtype match, then the rdata is compared
1378 bytewise until a difference is found. For example, in the common case
1379 of advertising DNS-SD services with a TXT record and an SRV record,
1380 the TXT record comes first (the rrtype for TXT is 16) and the SRV
1381 record comes second (the rrtype for SRV is 33).
1383 When comparing the records, if the first records match perfectly,
1384 then the second records are compared, and so on. If either list of
1385 records runs out of records before any difference is found, then the
1386 list with records remaining is deemed to have won the tie-break. If
1387 both lists run out of records at the same time without any difference
1388 being found, then this indicates that two devices are advertising
1389 identical sets of records, as is sometimes done for fault tolerance,
1390 and there is in fact no conflict.
1392 Expires 10th February 2007 Cheshire & Krochmal [Page 24]
1394 Internet Draft Multicast DNS 10th August 2006
1399 The second startup step is that the Multicast DNS Responder MUST send
1400 a gratuitous Multicast DNS Response containing, in the Answer
1401 Section, all of its resource records (both shared records, and unique
1402 records that have completed the probing step). If there are too many
1403 resource records to fit in a single packet, multiple packets should
1406 In the case of shared records (e.g. the PTR records used by DNS
1407 Service Discovery [DNS-SD]), the records are simply placed as-is
1408 into the Answer Section of the DNS Response.
1410 In the case of records that have been verified to be unique in the
1411 previous step, they are placed into the Answer Section of the DNS
1412 Response with the most significant bit of the rrclass set to one.
1413 The most significant bit of the rrclass for a record in the Answer
1414 Section of a response packet is the mDNS "cache flush" bit and is
1415 discussed in more detail below in Section 11.3 "Announcements to
1416 Flush Outdated Cache Entries".
1418 The Multicast DNS Responder MUST send at least two gratuitous
1419 responses, one second apart. A Responder MAY send up to eight
1420 gratuitous Responses, provided that the interval between gratuitous
1421 responses doubles with every response sent.
1423 A Multicast DNS Responder MUST NOT send announcements in the absence
1424 of information that its network connectivity may have changed in
1425 some relevant way. In particular, a Multicast DNS Responder MUST NOT
1426 send regular periodic announcements as a matter of course. It is not
1427 uncommon for protocol designers to encounter some problem which they
1428 decide to solve using regular periodic announcements, but this is
1429 generally not a wise protocol design choice. In the small scale
1430 periodic announcements may seem to remedy the short-term problem,
1431 but they do not scale well if the protocol becomes successful.
1432 If every host on the network implements the protocol -- if multiple
1433 applications on every host on the network are implementing the
1434 protocol -- then even a low periodic rate of just one announcement
1435 per minute per application per host can add up to multiple packets
1436 per second in total. While gigabit Ethernet may be able to carry
1437 a million packets per second, other network technologies cannot.
1438 For example, while IEEE 802.11g wireless has a nominal data rate of
1439 up to 54Mb/sec, multicasting just 100 packets per second can consume
1440 the entire available bandwidth, leaving nothing for anything else.
1442 With the increasing popularity of hand-held devices, unnecessary
1443 continuous packet transmission can have bad implications for battery
1444 life. It's worth pointing out the precedent that TCP was also
1445 designed with this "no regular periodic idle packets" philosophy.
1446 Standard TCP sends packets only when it has data to send or
1447 acknowledge. If neither client nor server sends any bytes, then the
1450 Expires 10th February 2007 Cheshire & Krochmal [Page 25]
1452 Internet Draft Multicast DNS 10th August 2006
1455 TCP code will send no packets, and a TCP connection can remain active
1456 in this state indefinitely, with no packets being exchanged for
1457 hours, days, weeks or months.
1459 Whenever a Multicast DNS Responder receives any Multicast DNS
1460 response (gratuitous or otherwise) containing a conflicting resource
1461 record, the conflict MUST be resolved as described below in "Conflict
1467 At any time, if the rdata of any of a host's Multicast DNS records
1468 changes, the host MUST repeat the Announcing step described above to
1469 update neighboring caches. For example, if any of a host's IP
1470 addresses change, it MUST re-announce those address records.
1472 In the case of shared records, a host MUST send a "goodbye"
1473 announcement with TTL zero (see Section 11.2 "Goodbye Packets")
1474 for the old rdata, to cause it to be deleted from peer caches,
1475 before announcing the new rdata. In the case of unique records,
1476 a host SHOULD omit the "goodbye" announcement, since the cache
1477 flush bit on the newly announced records will cause old rdata
1478 to be flushed from peer caches anyway.
1480 A host may update the contents of any of its records at any time,
1481 though a host SHOULD NOT update records more frequently than ten
1482 times per minute. Frequent rapid updates impose a burden on the
1483 network. If a host has information to disseminate which changes more
1484 frequently than ten times per minute, then it may be more appropriate
1485 to design a protocol for that specific purpose.
1488 10. Conflict Resolution
1490 A conflict occurs when a Multicast DNS Responder has a unique record
1491 for which it is authoritative, and it receives a Multicast DNS
1492 response packet containing a record with the same name, rrtype and
1493 rrclass, but inconsistent rdata. What may be considered inconsistent
1494 is context sensitive, except that resource records with identical
1495 rdata are never considered inconsistent, even if they originate from
1496 different hosts. This is to permit use of proxies and other
1497 fault-tolerance mechanisms that may cause more than one responder
1498 to be capable of issuing identical answers on the network.
1500 A common example of a resource record type that is intended to be
1501 unique, not shared between hosts, is the address record that maps a
1502 host's name to its IP address. Should a host witness another host
1503 announce an address record with the same name but a different IP
1504 address, then that is considered inconsistent, and that address
1505 record is considered to be in conflict.
1508 Expires 10th February 2007 Cheshire & Krochmal [Page 26]
1510 Internet Draft Multicast DNS 10th August 2006
1513 Whenever a Multicast DNS Responder receives any Multicast DNS
1514 response (gratuitous or otherwise) containing a conflicting resource
1515 record in the Answer Section, the Multicast DNS Responder MUST
1516 immediately reset its conflicted unique record to probing state, and
1517 go through the startup steps described above in Section 9. "Probing
1518 and Announcing on Startup". The protocol used in the Probing phase
1519 will determine a winner and a loser, and the loser MUST cease using
1520 the name, and reconfigure.
1522 It is very important that any host receiving a resource record that
1523 conflicts with one of its own MUST take action as described above.
1524 In the case of two hosts using the same host name, where one has been
1525 configured to require a unique host name and the other has not, the
1526 one that has not been configured to require a unique host name will
1527 not perceive any conflict, and will not take any action. By reverting
1528 to Probing state, the host that desires a unique host name will go
1529 through the necessary steps to ensure that a unique host is obtained.
1531 The recommended course of action after probing and failing is as
1534 o Programmatically change the resource record name in an attempt to
1535 find a new name that is unique. This could be done by adding some
1536 further identifying information (e.g. the model name of the
1537 hardware) if it is not already present in the name, appending the
1538 digit "2" to the name, or incrementing a number at the end of the
1539 name if one is already present.
1541 o Probe again, and repeat until a unique name is found.
1543 o Record this newly chosen name in persistent storage so that the
1544 device will use the same name the next time it is power-cycled.
1546 o Display a message to the user or operator informing them of the
1547 name change. For example:
1549 The name "Bob's Music" is in use by another iTunes music
1550 server on the network. Your music has been renamed to
1551 "Bob's Music (G4 Cube)". If you want to change this name,
1552 use [describe appropriate menu item or preference dialog].
1554 o If after one minute of probing the Multicast DNS Responder has been
1555 unable to find any unused name, it should display a message to the
1556 user or operator informing them of this fact. This situation should
1557 never occur in normal operation. The only situations that would
1558 cause this to happen would be either a deliberate denial-of-service
1559 attack, or some kind of very obscure hardware or software bug that
1560 acts like a deliberate denial-of-service attack.
1562 How the user or operator is informed depends on context. A desktop
1563 computer with a screen might put up a dialog box. A headless server
1566 Expires 10th February 2007 Cheshire & Krochmal [Page 27]
1568 Internet Draft Multicast DNS 10th August 2006
1571 in the closet may write a message to a log file, or use whatever
1572 mechanism (email, SNMP trap, etc.) it uses to inform the
1573 administrator of other error conditions. On the other hand a headless
1574 server in the closet may not inform the user at all -- if the user
1575 cares, they will notice the name has changed, and connect to the
1576 server in the usual way (e.g. via Web Browser) to configure a new
1579 The examples in this section focus on address records (i.e. host
1580 names), but the same considerations apply to all resource records
1581 where uniqueness (or maintenance of some other defined constraint)
1585 11. Resource Record TTL Values and Cache Coherency
1587 As a general rule, the recommended TTL value for Multicast DNS
1588 resource records with a host name as the resource record's name
1589 (e.g. A, AAAA, HINFO, etc.) or contained within the resource record's
1590 rdata (e.g. SRV, reverse mapping PTR record, etc.) is 120 seconds.
1592 The recommended TTL value for other Multicast DNS resource records
1595 A client with an active outstanding query will issue a query packet
1596 when one or more of the resource record(s) in its cache is (are) 80%
1597 of the way to expiry. If the TTL on those records is 75 minutes,
1598 this ongoing cache maintenance process yields a steady-state query
1599 rate of one query every 60 minutes.
1601 Any distributed cache needs a cache coherency protocol. If Multicast
1602 DNS resource records follow the recommendation and have a TTL of 75
1603 minutes, that means that stale data could persist in the system for
1604 a little over an hour. Making the default TTL significantly lower
1605 would reduce the lifetime of stale data, but would produce too much
1606 extra traffic on the network. Various techniques are available to
1607 minimize the impact of such stale data.
1610 11.1 Cooperating Multicast DNS Responders
1612 If a Multicast DNS Responder ("A") observes some other Multicast DNS
1613 Responder ("B") send a Multicast DNS Response packet containing a
1614 resource record with the same name, rrtype and rrclass as one of A's
1615 resource records, but different rdata, then:
1617 o If A's resource record is intended to be a shared resource record,
1618 then this is no conflict, and no action is required.
1620 o If A's resource record is intended to be a member of a unique
1621 resource record set owned solely by that responder, then this
1624 Expires 10th February 2007 Cheshire & Krochmal [Page 28]
1626 Internet Draft Multicast DNS 10th August 2006
1629 is a conflict and MUST be handled as described in Section 10
1630 "Conflict Resolution".
1632 If a Multicast DNS Responder ("A") observes some other Multicast DNS
1633 Responder ("B") send a Multicast DNS Response packet containing a
1634 resource record with the same name, rrtype and rrclass as one of A's
1635 resource records, and identical rdata, then:
1637 o If the TTL of B's resource record given in the packet is at least
1638 half the true TTL from A's point of view, then no action is
1641 o If the TTL of B's resource record given in the packet is less than
1642 half the true TTL from A's point of view, then A MUST mark its
1643 record to be announced via multicast. Clients receiving the record
1644 from B would use the TTL given by B, and hence may delete the
1645 record sooner than A expects. By sending its own multicast response
1646 correcting the TTL, A ensures that the record will be retained for
1649 These rules allow multiple Multicast DNS Responders to offer the same
1650 data on the network (perhaps for fault tolerance reasons) without
1651 conflicting with each other.
1654 11.2 Goodbye Packets
1656 In the case where a host knows that certain resource record data is
1657 about to become invalid (for example when the host is undergoing a
1658 clean shutdown) the host SHOULD send a gratuitous announcement mDNS
1659 response packet, giving the same resource record name, rrtype,
1660 rrclass and rdata, but an RR TTL of zero. This has the effect of
1661 updating the TTL stored in neighboring hosts' cache entries to zero,
1662 causing that cache entry to be promptly deleted.
1664 Clients receiving a Multicast DNS Response with a TTL of zero SHOULD
1665 NOT immediately delete the record from the cache, but instead record
1666 a TTL of 1 and then delete the record one second later. In the case
1667 of multiple Multicast DNS Responders on the network described in
1668 Section 11.1 above, if one of the Responders shuts down and
1669 incorrectly sends goodbye packets for its records, it gives the other
1670 cooperating Responders one second to send out their own response to
1671 "rescue" the records before they expire and are deleted.
1674 11.3 Announcements to Flush Outdated Cache Entries
1676 Whenever a host has a resource record with potentially new data (e.g.
1677 after rebooting, waking from sleep, connecting to a new network link,
1678 changing IP address, etc.), the host MUST send a series of gratuitous
1679 announcements to update cache entries in its neighbor hosts. In
1682 Expires 10th February 2007 Cheshire & Krochmal [Page 29]
1684 Internet Draft Multicast DNS 10th August 2006
1687 these gratuitous announcements, if the record is one that is intended
1688 to be unique, the host sets the most significant bit of the rrclass
1689 field of the resource record. This bit, the "cache flush" bit, tells
1690 neighboring hosts that this is not a shared record type. Instead of
1691 merging this new record additively into the cache in addition to any
1692 previous records with the same name, rrtype and rrclass, all old
1693 records with that name, type and class that were received more than
1694 one second ago are declared invalid, and marked to expire from the
1695 cache in one second.
1697 The semantics of the cache flush bit are as follows: Normally when a
1698 resource record appears in the Answer Section of the DNS Response, it
1699 means, "This is an assertion that this information is true." When a
1700 resource record appears in the Answer Section of the DNS Response
1701 with the "cache flush" bit set, it means, "This is an assertion that
1702 this information is the truth and the whole truth, and anything you
1703 may have heard more than a second ago regarding records of this
1704 name/rrtype/rrclass is no longer valid".
1706 To accommodate the case where the set of records from one host
1707 constituting a single unique RRSet is too large to fit in a single
1708 packet, only cache records that are more than one second old are
1709 flushed. This allows the announcing host to generate a quick burst of
1710 packets back-to-back on the wire containing all the members
1711 of the RRSet. When receiving records with the "cache flush" bit set,
1712 all records older than one second are marked to be deleted one second
1713 in the future. One second after the end of the little packet burst,
1714 any records not represented within that packet burst will then be
1715 expired from all peer caches.
1717 Any time a host sends a response packet containing some members of a
1718 unique RRSet, it SHOULD send the entire RRSet, preferably in a single
1719 packet, or if the entire RRSet will not fit in a single packet, in a
1720 quick burst of packets sent as close together as possible. The host
1721 SHOULD set the cache flush bit on all members of the unique RRSet.
1722 In the event that for some reason the host chooses not to send the
1723 entire unique RRSet in a single packet or a rapid packet burst,
1724 it MUST NOT set the cache flush bit on any of those records.
1726 The reason for waiting one second before deleting stale records from
1727 the cache is to accommodate bridged networks. For example, a host's
1728 address record announcement on a wireless interface may be bridged
1729 onto a wired Ethernet, and cause that same host's Ethernet address
1730 records to be flushed from peer caches. The one-second delay gives
1731 the host the chance to see its own announcement arrive on the wired
1732 Ethernet, and immediately re-announce its Ethernet interface's
1733 address records so that both sets remain valid and live in peer
1736 These rules apply regardless of *why* the response packet is being
1737 generated. They apply to startup announcements as described in
1740 Expires 10th February 2007 Cheshire & Krochmal [Page 30]
1742 Internet Draft Multicast DNS 10th August 2006
1745 Section 9.3 "Announcing", and to responses generated as a result
1746 of receiving query packets.
1748 The "cache flush" bit is only set in records in the Answer Section of
1749 Multicast DNS responses sent to UDP port 5353. The "cache flush" bit
1750 MUST NOT be set in any resource records in a response packet sent in
1751 legacy unicast responses to UDP ports other than 5353.
1753 The "cache flush" bit MUST NOT be set in any resource records in the
1754 known-answer list of any query packet.
1756 The "cache flush" bit MUST NOT ever be set in any shared resource
1757 record. To do so would cause all the other shared versions of this
1758 resource record with different rdata from different Responders to be
1759 immediately deleted from all the caches on the network.
1761 The "cache flush" bit does apply to questions listed in the Question
1762 Section of a Multicast DNS packet. The top bit of the rrclass field
1763 in questions is used for an entirely different purpose (see Section
1764 6.5, "Questions Requesting Unicast Responses").
1766 Note that the "cache flush" bit is NOT part of the resource record
1767 class. The "cache flush" bit is the most significant bit of the
1768 second 16-bit word of a resource record in the Answer Section of
1769 an mDNS packet (the field conventionally referred to as the rrclass
1770 field), and the actual resource record class is the least-significant
1771 fifteen bits of this field. There is no mDNS resource record class
1772 0x8001. The value 0x8001 in the rrclass field of a resource record in
1773 an mDNS response packet indicates a resource record with class 1,
1774 with the "cache flush" bit set. When receiving a resource record with
1775 the "cache flush" bit set, implementations should take care to mask
1776 off that bit before storing the resource record in memory.
1779 11.4 Cache Flush on Topology change
1781 If the hardware on a given host is able to indicate physical changes
1782 of connectivity, then when the hardware indicates such a change, the
1783 host should take this information into account in its mDNS cache
1784 management strategy. For example, a host may choose to immediately
1785 flush all cache records received on a particular interface when that
1786 cable is disconnected. Alternatively, a host may choose to adjust the
1787 remaining TTL on all those records to a few seconds so that if the
1788 cable is not reconnected quickly, those records will expire from the
1791 Likewise, when a host reboots, or wakes from sleep, or undergoes some
1792 other similar discontinuous state change, the cache management
1793 strategy should take that information into account.
1798 Expires 10th February 2007 Cheshire & Krochmal [Page 31]
1800 Internet Draft Multicast DNS 10th August 2006
1803 11.5 Cache Flush on Failure Indication
1805 Sometimes a cache record can be determined to be stale when a client
1806 attempts to use the rdata it contains, and finds that rdata to be
1809 For example, the rdata in an address record can be determined to be
1810 incorrect if attempts to contact that host fail, either because
1811 ARP/ND requests for that address go unanswered (for an address on a
1812 local subnet) or because a router returns an ICMP "Host Unreachable"
1813 error (for an address on a remote subnet).
1815 The rdata in an SRV record can be determined to be incorrect if
1816 attempts to communicate with the indicated service at the host and
1817 port number indicated are not successful.
1819 The rdata in a DNS-SD PTR record can be determined to be incorrect if
1820 attempts to look up the SRV record it references are not successful.
1822 In any such case, the software implementing the mDNS resource record
1823 cache should provide a mechanism so that clients detecting stale
1824 rdata can inform the cache.
1826 When the cache receives this hint that it should reconfirm some
1827 record, it MUST issue two or more queries for the resource record in
1828 question. If no response is received in a reasonable amount of time,
1829 then, even though its TTL may indicate that it is not yet due to
1830 expire, that record SHOULD be promptly flushed from the cache.
1832 The end result of this is that if a printer suffers a sudden power
1833 failure or other abrupt disconnection from the network, its name
1834 may continue to appear in DNS-SD browser lists displayed on users'
1835 screens. Eventually that entry will expire from the cache naturally,
1836 but if a user tries to access the printer before that happens, the
1837 failure to successfully contact the printer will trigger the more
1838 hasty demise of its cache entries. This is a sensible trade-off
1839 between good user-experience and good network efficiency. If we were
1840 to insist that printers should disappear from the printer list within
1841 30 seconds of becoming unavailable, for all failure modes, the only
1842 way to achieve this would be for the client to poll the printer at
1843 least every 30 seconds, or for the printer to announce its presence
1844 at least every 30 seconds, both of which would be an unreasonable
1845 burden on most networks.
1848 11.6 Passive Observation of Failures
1850 A host observes the multicast queries issued by the other hosts on
1851 the network. One of the major benefits of also sending responses
1852 using multicast is that it allows all hosts to see the responses (or
1853 lack thereof) to those queries.
1856 Expires 10th February 2007 Cheshire & Krochmal [Page 32]
1858 Internet Draft Multicast DNS 10th August 2006
1861 If a host sees queries, for which a record in its cache would be
1862 expected to be given as an answer in a multicast response, but no
1863 such answer is seen, then the host may take this as an indication
1864 that the record may no longer be valid.
1866 After seeing two or more of these queries, and seeing no multicast
1867 response containing the expected answer within a reasonable amount of
1868 time, then even though its TTL may indicate that it is not yet due to
1869 expire, that record MAY be flushed from the cache. The host SHOULD
1870 NOT perform its own queries to re-confirm that the record is truly
1871 gone. If every host on a large network were to do this, it would
1872 cause a lot of unnecessary multicast traffic. If host A sends
1873 multicast queries that remain unanswered, then there is no reason
1874 to suppose that host B or any other host is likely to be any more
1877 The previous section, "Cache Flush on Failure Indication", describes
1878 a situation where a user trying to print discovers that the printer
1879 is no longer available. By implementing the passive observation
1880 described here, when one user fails to contact the printer, all
1881 hosts on the network observe that failure and update their caches
1885 12. Special Characteristics of Multicast DNS Domains
1887 Unlike conventional DNS names, names that end in ".local." or
1888 "254.169.in-addr.arpa." have only local significance. The same is
1889 true of names within the IPv6 Link-Local reverse mapping domains.
1891 Conventional Unicast DNS seeks to provide a single unified namespace,
1892 where a given DNS query yields the same answer no matter where on the
1893 planet it is performed or to which recursive DNS server the query is
1894 sent. In contrast, each IP link has its own private ".local.",
1895 "254.169.in-addr.arpa." and IPv6 Link-Local reverse mapping
1896 namespaces, and the answer to any query for a name within those
1897 domains depends on where that query is asked. (This characteristic is
1898 not unique to Multicast DNS. Although the original concept of DNS was
1899 a single global namespace, in recent years split views, firewalls,
1900 intranets, and the like have increasingly meant that the answer to a
1901 given DNS query has become dependent on the location of the querier.)
1903 Multicast DNS Domains are not delegated from their parent domain via
1904 use of NS records. There are no NS records anywhere in Multicast DNS
1905 Domains. Instead, all Multicast DNS Domains are delegated to the IP
1906 addresses 224.0.0.251 and FF02::FB by virtue of the individual
1907 organizations producing DNS client software deciding how to handle
1908 those names. It would be extremely valuable for the industry if this
1909 special handling were ratified and recorded by IANA, since otherwise
1910 the special handling provided by each vendor is likely to be
1914 Expires 10th February 2007 Cheshire & Krochmal [Page 33]
1916 Internet Draft Multicast DNS 10th August 2006
1919 The IPv4 name server for a Multicast DNS Domain is 224.0.0.251. The
1920 IPv6 name server for a Multicast DNS Domain is FF02::FB. These are
1921 multicast addresses; therefore they identify not a single host but a
1922 collection of hosts, working in cooperation to maintain some
1923 reasonable facsimile of a competently managed DNS zone. Conceptually
1924 a Multicast DNS Domain is a single DNS zone, however its server is
1925 implemented as a distributed process running on a cluster of loosely
1926 cooperating CPUs rather than as a single process running on a single
1929 No delegation is performed within Multicast DNS Domains. Because the
1930 cluster of loosely coordinated CPUs is cooperating to administer a
1931 single zone, delegation is neither necessary nor desirable. Just
1932 because a particular host on the network may answer queries for a
1933 particular record type with the name "example.local." does not imply
1934 anything about whether that host will answer for the name
1935 "child.example.local.", or indeed for other record types with the
1936 name "example.local."
1938 Multicast DNS Zones have no SOA record. A conventional DNS zone's
1939 SOA record contains information such as the email address of the zone
1940 administrator and the monotonically increasing serial number of the
1941 last zone modification. There is no single human administrator for
1942 any given Multicast DNS Zone, so there is no email address. Because
1943 the hosts managing any given Multicast DNS Zone are only loosely
1944 coordinated, there is no readily available monotonically increasing
1945 serial number to determine whether or not the zone contents have
1946 changed. A host holding part of the shared zone could crash or be
1947 disconnected from the network at any time without informing the other
1948 hosts. There is no reliable way to provide a zone serial number that
1949 would, whenever such a crash or disconnection occurred, immediately
1950 change to indicate that the contents of the shared zone had changed.
1952 Zone transfers are not possible for any Multicast DNS Zone.
1955 13. Multicast DNS for Service Discovery
1957 This document does not describe using Multicast DNS for network
1958 browsing or service discovery. However, the mechanisms this document
1959 describes are compatible with (and support) the browsing and service
1960 discovery mechanisms proposed in "DNS-Based Service Discovery"
1964 14. Enabling and Disabling Multicast DNS
1966 The option to fail-over to Multicast DNS for names not ending
1967 in ".local." SHOULD be a user-configured option, and SHOULD
1968 be disabled by default because of the possible security issues
1969 related to unintended local resolution of apparently global names.
1972 Expires 10th February 2007 Cheshire & Krochmal [Page 34]
1974 Internet Draft Multicast DNS 10th August 2006
1977 The option to lookup unqualified (relative) names by appending
1978 ".local." (or not) is controlled by whether ".local." appears
1979 (or not) in the client's DNS search list.
1981 No special control is needed for enabling and disabling Multicast DNS
1982 for names explicitly ending with ".local." as entered by the user.
1983 The user doesn't need a way to disable Multicast DNS for names ending
1984 with ".local.", because if the user doesn't want to use Multicast
1985 DNS, they can achieve this by simply not using those names. If a user
1986 *does* enter a name ending in ".local.", then we can safely assume
1987 the user's intention was probably that it should work. Having user
1988 configuration options that can be (intentionally or unintentionally)
1989 set so that local names don't work is just one more way of
1990 frustrating the user's ability to perform the tasks they want,
1991 perpetuating the view that, "IP networking is too complicated to
1992 configure and too hard to use." This in turn perpetuates the
1993 continued use of protocols like AppleTalk. If we want to retire
1994 AppleTalk, NetBIOS, etc., we need to offer users equivalent IP
1995 functionality that they can rely on to, "always work, like
1996 AppleTalk." A little Multicast DNS traffic may be a burden on the
1997 network, but it is an insignificant burden compared to continued
1998 widespread use of AppleTalk.
2001 15. Considerations for Multiple Interfaces
2003 A host SHOULD defend its host name (FQDN) on all active interfaces on
2004 which it is answering Multicast DNS queries.
2006 In the event of a name conflict on *any* interface, a host should
2007 configure a new host name, if it wishes to maintain uniqueness of its
2010 A host may choose to use the same name for all of its address records
2011 on all interfaces, or it may choose to manage its Multicast DNS host
2012 name(s) independently on each interface, potentially answering to
2013 different names on different interfaces.
2015 When answering a Multicast DNS query, a multi-homed host with a
2016 link-local address (or addresses) SHOULD take care to ensure that
2017 any address going out in a Multicast DNS response is valid for use
2018 on the interface on which the response is going out.
2020 Just as the same link-local IP address may validly be in use
2021 simultaneously on different links by different hosts, the same
2022 link-local host name may validly be in use simultaneously on
2023 different links, and this is not an error. A multi-homed host with
2024 connections to two different links may be able to communicate with
2025 two different hosts that are validly using the same name. While this
2026 kind of name duplication should be rare, it means that a host that
2027 wants to fully support this case needs network programming APIs that
2028 allow applications to specify on what interface to perform a
2030 Expires 10th February 2007 Cheshire & Krochmal [Page 35]
2032 Internet Draft Multicast DNS 10th August 2006
2035 link-local Multicast DNS query, and to discover on what interface a
2036 Multicast DNS response was received.
2038 There is one other special precaution that multi-homed hosts need to
2039 take. It's common with today's laptop computers to have an Ethernet
2040 connection and an 802.11 wireless connection active at the same time.
2041 What the software on the laptop computer can't easily tell is whether
2042 the wireless connection is in fact bridged onto the same network
2043 segment as its Ethernet connection. If the two networks are bridged
2044 together, then packets the host sends on one interface will arrive on
2045 the other interface a few milliseconds later, and care must be taken
2046 to ensure that this bridging does not cause problems:
2048 When the host announces its host name (i.e. its address records) on
2049 its wireless interface, those announcement records are sent with the
2050 cache-flush bit set, so when they arrive on the Ethernet segment,
2051 they will cause all the peers on the Ethernet to flush the host's
2052 Ethernet address records from their caches. The mDNS protocol has a
2053 safeguard to protect against this situation: when records are
2054 received with the cache-flush bit set, other records are not deleted
2055 from peer caches immediately, but are marked for deletion in one
2056 second. When the host sees its own wireless address records arrive on
2057 its Ethernet interface, with the cache-flush bit set, this one-second
2058 grace period gives the host time to respond and re-announce its
2059 Ethernet address records, to reinstate those records in peer caches
2060 before they are deleted.
2062 As described, this solves one problem, but creates another, because
2063 when those Ethernet announcement records arrive back on the wireless
2064 interface, the host would again respond defensively to reinstate its
2065 wireless records, and this process would continue forever,
2066 continuously flooding the network with traffic. The mDNS protocol has
2067 a second safeguard, to solve this problem: the cache-flush bit does
2068 not apply to records received very recently, within the last second.
2069 This means that when the host sees its own Ethernet address records
2070 arrive on its wireless interface, with the cache-flush bit set, it
2071 knows there's no need to re-announce its wireless address records
2072 again because it already sent them less than a second ago, and this
2073 makes them immune from deletion from peer caches.
2075 16. Considerations for Multiple Responders on the Same Machine
2077 It is possible to have more than one Multicast DNS Responder and/or
2078 Querier implementation coexist on the same machine, but there are
2081 16.1 Receiving Unicast Responses
2083 In most operating systems, incoming multicast packets can be
2084 delivered to *all* open sockets bound to the right port number,
2085 provided that the clients take the appropriate steps to allow this.
2086 For this reason, all Multicast DNS implementations SHOULD use the
2088 Expires 10th February 2007 Cheshire & Krochmal [Page 36]
2090 Internet Draft Multicast DNS 10th August 2006
2093 SO_REUSEPORT and/or SO_REUSEADDR options (or equivalent as
2094 appropriate for the operating system in question) so they will all be
2095 able to bind to UDP port 5353 and receive incoming multicast packets
2096 addressed to that port. However, incoming unicast UDP packets are
2097 typically delivered only to the first socket to bind to that port.
2098 This means that "QU" responses and other packets sent via unicast
2099 will be received only by the first Multicast DNS Responder and/or
2100 Querier on a system. This limitation can be partially mitigated if
2101 Multicast DNS implementations detect when they are not the first
2102 to bind to port 5353, and in that case they do not request "QU"
2103 responses. One way to detect if there is another Multicast DNS
2104 implementation already running is to attempt binding to port 5353
2105 without using SO_REUSEPORT and/or SO_REUSEADDR, and if that fails
2106 it indicates that some other socket is already bound to this port.
2109 16.2 Multi-Packet Known-Answer lists
2111 When a Multicast DNS Querier issues a query with too many known
2112 answers to fit into a single packet, it divides the known answer list
2113 into two or more packets. Multicast DNS Responders associate the
2114 initial truncated query with its continuation packets by examining
2115 the source IP address in each packet. Since two independent Multicast
2116 DNS Queriers running on the same machine will be sending packets with
2117 the same source IP address, from an outside perspective they appear
2118 to be a single entity. If both Queriers happened to send the same
2119 multi-packet query at the same time, with different known answer
2120 lists, then they could each end up suppressing answers that the other
2126 If different clients on a machine were to each have their own
2127 separate independent Multicast DNS implementation, they would lose
2128 certain efficiency benefits. Apart from the unnecessary code
2129 duplication, memory usage, and CPU load, the clients wouldn't get the
2130 benefit of a shared system-wide cache, and they would not be able to
2131 aggregate separate queries into single packets to reduce network
2137 Because of these issues, this document encourages implementers
2138 to design systems with a single Multicast DNS implementation that
2139 provides Multicast DNS services shared by all clients on that
2140 machine. Due to engineering constraints, there may be situations
2141 where embedding a Multicast DNS implementation in the client is the
2142 most expedient solution, and while this will work in practice,
2143 implementers should be aware of the issues outlined in this section.
2146 Expires 10th February 2007 Cheshire & Krochmal [Page 37]
2148 Internet Draft Multicast DNS 10th August 2006
2150 17. Multicast DNS and Power Management
2152 Many modern network devices have the ability to go into a low-power
2153 mode where only a small part of the Ethernet hardware remains
2154 powered, and the device can be woken up by sending a specially
2155 formatted Ethernet frame which the device's power-management hardware
2158 To make use of this in conjunction with Multicast DNS, we propose a
2159 network power management service called Sleep Proxy Service. A device
2160 that wishes to enter low-power mode first uses DNS-SD to determine if
2161 Sleep Proxy Service is available on the local network. In some
2162 networks there may be more than one piece of hardware implementing
2163 Sleep Proxy Service, for fault-tolerance reasons.
2165 If the device finds the network has Sleep Proxy Service, the device
2166 transmits two or more gratuitous mDNS announcements setting the TTL
2167 of its relevant resource records to zero, to delete them from
2168 neighboring caches. The relevant resource records include address
2169 records and SRV records, and other resource records as may apply to a
2170 particular device. The device then communicates all of its remaining
2171 active records, plus the names, rrtypes and rrclasses of the deleted
2172 records, to the Sleep Proxy Service(s), along with a copy of the
2173 specific "magic packet" required to wake the device up.
2175 When a Sleep Proxy Service sees an mDNS query for one of the
2176 device's active records (e.g. a DNS-SD PTR record), it answers on
2177 behalf of the device without waking it up. When a Sleep Proxy Service
2178 sees an mDNS query for one of the device's deleted resource
2179 records, it deduces that some client on the network needs to make an
2180 active connection to the device, and sends the specified "magic
2181 packet" to wake the device up. The device then wakes up, reactivates
2182 its deleted resource records, and re-announces them to the network.
2183 The client waiting to connect sees the announcements, learns the
2184 current IP address and port number of the desired service on the
2185 device, and proceeds to connect to it.
2187 The connecting client does not need to be aware of how Sleep Proxy
2188 Service works. Only devices that implement low power mode and wish to
2189 make use of Sleep Proxy Service need to be aware of how that protocol
2192 The reason that a device using a Sleep Proxy Service should send more
2193 than one goodbye packet is to ensure deletion of the resource records
2194 from all peer caches. If resource records were to inadvertently
2195 remain in some peer caches, then those peers may not issue any query
2196 packets for those records when attempting to access the sleeping
2197 device, so the Sleep Proxy Service would not receive any queries for
2198 the device's SRV and/or address records, and the necessary wake-up
2199 message would not be triggered.
2201 The full specification of mDNS / DNS-SD Sleep Proxy Service
2202 is described in another document [not yet published].
2204 Expires 10th February 2007 Cheshire & Krochmal [Page 38]
2206 Internet Draft Multicast DNS 10th August 2006
2209 18. Multicast DNS Character Set
2211 Unicast DNS has been plagued by the lack of any support for non-US
2212 characters. Indeed, conventional DNS is usually limited to just
2213 letters, digits and hyphens, with no spaces or other punctuation.
2214 Attempts to remedy this for unicast DNS have been badly constrained
2215 by the need to accommodate old buggy legacy DNS implementations.
2216 In reality, the DNS specification actually imposes no limits on what
2217 characters may be used in names, and good DNS implementations handle
2218 any arbitrary eight-bit data without trouble. However, the old rules
2219 for ARPANET host names back in the 1980s required names to be just
2220 letters, digits, and hyphens [RFC 1034], and since the predominant
2221 use of DNS is to store host address records, many have assumed that
2222 the DNS protocol itself suffers from the same limitation. It would be
2223 more accurate to say that certain bad implementations may not handle
2224 eight-bit data correctly, not that the protocol doesn't support it.
2226 Multicast DNS is a new protocol and doesn't (yet) have old buggy
2227 legacy implementations to constrain the design choices. Accordingly,
2228 it adopts the simple obvious elegant solution: all names in Multicast
2229 DNS are encoded using precomposed UTF-8 [RFC 3629]. The characters
2230 SHOULD conform to Unicode Normalization Form C (NFC) [UAX15]: Use
2231 precomposed characters instead of combining sequences where possible,
2232 e.g. use U+00C4 ("Latin capital letter A with diaeresis") instead of
2233 U+0041 U+0308 ("Latin capital letter A", "combining diaeresis").
2235 Some users of 16-bit Unicode have taken to stuffing a "zero-width
2236 non-breaking space" character (U+FEFF) at the start of each UTF-16
2237 file, as a hint to identify whether the data is big-endian or
2238 little-endian, and calling it a "Byte Order Mark" (BOM). Since there
2239 is only one possible byte order for UTF-8 data, a BOM is neither
2240 necessary nor permitted. Multicast DNS names MUST NOT contain a "Byte
2241 Order Mark". Any occurrence of the Unicode character U+FEFF at the
2242 start or anywhere else in a Multicast DNS name MUST be interpreted as
2243 being an actual intended part of the name, representing (just as for
2244 any other legal unicode value) an actual literal instance of that
2245 character (in this case a zero-width non-breaking space character).
2247 For names that are restricted to letters, digits and hyphens, the
2248 UTF-8 encoding is identical to the US-ASCII encoding, so this is
2249 entirely compatible with existing host names. For characters outside
2250 the US-ASCII range, UTF-8 encoding is used.
2252 Multicast DNS implementations MUST NOT use any other encodings apart
2253 from precomposed UTF-8 (US-ASCII being considered a compatible subset
2256 This point bears repeating: After many years of debate, as a
2257 result of the need to accommodate certain DNS implementations that
2258 apparently couldn't handle any character that's not a letter, digit
2259 or hyphen (and apparently never will be updated to remedy this
2262 Expires 10th February 2007 Cheshire & Krochmal [Page 39]
2264 Internet Draft Multicast DNS 10th August 2006
2267 limitation) the unicast DNS community settled on an extremely baroque
2268 encoding called "Punycode" [RFC 3492]. Punycode is a remarkably
2269 ingenious encoding solution, but it is complicated, hard to
2270 understand, and hard to implement, using sophisticated techniques
2271 including insertion unsort coding, generalized variable-length
2272 integers, and bias adaptation. The resulting encoding is remarkably
2273 compact given the constraints, but it's still not as good as simple
2274 straightforward UTF-8, and it's hard even to predict whether a given
2275 input string will encode to a Punycode string that fits within DNS's
2276 63-byte limit, except by simply trying the encoding and seeing
2277 whether it fits. Indeed, the encoded size depends not only on the
2278 input characters, but on the order they appear, so the same set of
2279 characters may or may not encode to a legal Punycode string that fits
2280 within DNS's 63-byte limit, depending on the order the characters
2281 appear. This is extremely hard to present in a user interface that
2282 explains to users why one name is allowed, but another name
2283 containing the exact same characters is not. Neither Punycode nor any
2284 other of the "Ascii Compatible Encodings" proposed for Unicast DNS
2285 may be used in Multicast DNS packets. Any text being represented
2286 internally in some other representation MUST be converted to
2287 canonical precomposed UTF-8 before being placed in any Multicast DNS
2290 The simple rules for case-insensitivity in Unicast DNS also apply in
2291 Multicast DNS; that is to say, in name comparisons, the lower-case
2292 letters "a" to "z" (0x61 to 0x7A) match their upper-case equivalents
2293 "A" to "Z" (0x41 to 0x5A). Hence, if a client issues a query for an
2294 address record with the name "cheshire.local", then a responder
2295 having an address record with the name "Cheshire.local" should
2296 issue a response. No other automatic equivalences should be assumed.
2297 In particular all UTF-8 multi-byte characters (codes 0x80 and higher)
2298 are compared by simple binary comparison of the raw byte values.
2299 Accented characters are *not* defined to be automatically equivalent
2300 to their unaccented counterparts. Where automatic equivalences are
2301 desired, this may be achieved through the use of programmatically-
2302 generated CNAME records. For example, if a responder has an address
2303 record for an accented name Y, and a client issues a query for a name
2304 X, where X is the same as Y with all the accents removed, then the
2305 responder may issue a response containing two resource records:
2306 A CNAME record "X CNAME Y", asserting that the requested name X
2307 (unaccented) is an alias for the true (accented) name Y, followed
2308 by the address record for Y.
2320 Expires 10th February 2007 Cheshire & Krochmal [Page 40]
2322 Internet Draft Multicast DNS 10th August 2006
2325 19. Multicast DNS Message Size
2327 RFC 1035 restricts DNS Messages carried by UDP to no more than 512
2328 bytes (not counting the IP or UDP headers) [RFC 1035]. For UDP
2329 packets carried over the wide-area Internet in 1987, this was
2330 appropriate. For link-local multicast packets on today's networks,
2331 there is no reason to retain this restriction. Given that the packets
2332 are by definition link-local, there are no Path MTU issues to
2335 Multicast DNS Messages carried by UDP may be up to the IP MTU of the
2336 physical interface, less the space required for the IP header (20
2337 bytes for IPv4; 40 bytes for IPv6) and the UDP header (8 bytes).
2339 In the case of a single mDNS Resource Record which is too large to
2340 fit in a single MTU-sized multicast response packet, a Multicast DNS
2341 Responder SHOULD send the Resource Record alone, in a single IP
2342 datagram, sent using multiple IP fragments. Resource Records this
2343 large SHOULD be avoided, except in the very rare cases where they
2344 really are the appropriate solution to the problem at hand.
2345 Implementers should be aware that many simple devices do not
2346 re-assemble fragmented IP datagrams, so large Resource Records
2347 SHOULD NOT be used except in specialized cases where the implementer
2348 knows that all receivers implement reassembly.
2350 A Multicast DNS packet larger than the interface MTU, which is sent
2351 using fragments, MUST NOT contain more than one Resource Record.
2353 Even when fragmentation is used, a Multicast DNS packet, including IP
2354 and UDP headers, MUST NOT exceed 9000 bytes.
2378 Expires 10th February 2007 Cheshire & Krochmal [Page 41]
2380 Internet Draft Multicast DNS 10th August 2006
2383 20. Multicast DNS Message Format
2385 This section describes specific restrictions on the allowable
2386 values for the header fields of a Multicast DNS message.
2389 20.1 ID (Query Identifier)
2391 Multicast DNS clients SHOULD listen for gratuitous responses
2392 issued by hosts booting up (or waking up from sleep or otherwise
2393 joining the network). Since these gratuitous responses may contain a
2394 useful answer to a question for which the client is currently
2395 awaiting an answer, Multicast DNS clients SHOULD examine all received
2396 Multicast DNS response messages for useful answers, without regard to
2397 the contents of the ID field or the Question Section. In Multicast
2398 DNS, knowing which particular query message (if any) is responsible
2399 for eliciting a particular response message is less interesting than
2400 knowing whether the response message contains useful information.
2402 Multicast DNS clients MAY cache any or all Multicast DNS response
2403 messages they receive, for possible future use, provided of course
2404 that normal TTL aging is performed on these cached resource records.
2406 In multicast query messages, the Query ID SHOULD be set to zero on
2409 In multicast responses, including gratuitous multicast responses, the
2410 Query ID MUST be set to zero on transmission, and MUST be ignored on
2413 In unicast response messages generated specifically in response to a
2414 particular (unicast or multicast) query, the Query ID MUST match the
2415 ID from the query message.
2418 20.2 QR (Query/Response) Bit
2420 In query messages, MUST be zero.
2421 In response messages, MUST be one.
2426 In both multicast query and multicast response messages, MUST be zero
2427 (only standard queries are currently supported over multicast, unless
2428 other queries are allowed by future IETF Standards Action).
2436 Expires 10th February 2007 Cheshire & Krochmal [Page 42]
2438 Internet Draft Multicast DNS 10th August 2006
2441 20.4 AA (Authoritative Answer) Bit
2443 In query messages, the Authoritative Answer bit MUST be zero on
2444 transmission, and MUST be ignored on reception.
2446 In response messages for Multicast Domains, the Authoritative Answer
2447 bit MUST be set to one (not setting this bit implies there's some
2448 other place where "better" information may be found) and MUST be
2449 ignored on reception.
2452 20.5 TC (Truncated) Bit
2454 In query messages, if the TC bit is set, it means that additional
2455 Known Answer records may be following shortly. A responder MAY choose
2456 to record this fact, and wait for those additional Known Answer
2457 records, before deciding whether to respond. If the TC bit is clear,
2458 it means that the querying host has no additional Known Answers.
2460 In multicast response messages, the TC bit MUST be zero on
2461 transmission, and MUST be ignored on reception.
2463 In legacy unicast response messages, the TC bit has the same meaning
2464 as in conventional unicast DNS: it means that the response was too
2465 large to fit in a single packet, so the client SHOULD re-issue its
2466 query using TCP in order to receive the larger response.
2469 20.6 RD (Recursion Desired) Bit
2471 In both multicast query and multicast response messages, the
2472 Recursion Desired bit SHOULD be zero on transmission, and MUST be
2473 ignored on reception.
2476 20.7 RA (Recursion Available) Bit
2478 In both multicast query and multicast response messages, the
2479 Recursion Available bit MUST be zero on transmission, and MUST be
2480 ignored on reception.
2485 In both query and response messages, the Zero bit MUST be zero on
2486 transmission, and MUST be ignored on reception.
2494 Expires 10th February 2007 Cheshire & Krochmal [Page 43]
2496 Internet Draft Multicast DNS 10th August 2006
2499 20.9 AD (Authentic Data) Bit [RFC 2535]
2501 In query messages the Authentic Data bit MUST be zero on
2502 transmission, and MUST be ignored on reception.
2504 In response messages, the Authentic Data bit MAY be set. Resolvers
2505 receiving response messages with the AD bit set MUST NOT trust the AD
2506 bit unless they trust the source of the message and either have a
2507 secure path to it or use DNS transaction security.
2510 20.10 CD (Checking Disabled) Bit [RFC 2535]
2512 In query messages, a resolver willing to do cryptography SHOULD set
2513 the Checking Disabled bit to permit it to impose its own policies.
2515 In response messages, the Checking Disabled bit MUST be zero on
2516 transmission, and MUST be ignored on reception.
2519 20.11 RCODE (Response Code)
2521 In both multicast query and multicast response messages, the Response
2522 Code MUST be zero on transmission. Multicast DNS messages received
2523 with non-zero Response Codes MUST be silently ignored.
2526 20.12 Repurposing of top bit of qclass in Question Section
2528 In the Question Section of a Multicast DNS Query, the top bit of the
2529 qclass field is used to indicate that unicast responses are preferred
2530 for this particular question.
2533 20.13 Repurposing of top bit of rrclass in Answer Section
2535 In the Answer Section of a Multicast DNS Response, the top bit of the
2536 rrclass field is used to indicate that the record is a member of a
2537 unique RRSet, and the entire RRSet has been sent together (in the
2538 same packet, or in consecutive packets if there are too many records
2539 to fit in a single packet).
2552 Expires 10th February 2007 Cheshire & Krochmal [Page 44]
2554 Internet Draft Multicast DNS 10th August 2006
2557 21. Choice of UDP Port Number
2559 Arguments were made for and against using Multicast on UDP port 53.
2560 The final decision was to use UDP port 5353. Some of the arguments
2561 for and against are given below.
2564 21.1 Arguments for using UDP port 53:
2566 * This is "just DNS", so it should be the same port.
2568 * There is less work to be done updating old clients to do simple
2569 mDNS queries. Only the destination address need be changed.
2570 In some cases, this can be achieved without any code changes,
2571 just by adding the address 224.0.0.251 to a configuration file.
2574 21.2 Arguments for using a different port (UDP port 5353):
2576 * This is not "just DNS". This is a DNS-like protocol, but different.
2578 * Changing client code to use a different port number is not hard.
2580 * Using the same port number makes it hard to run an mDNS Responder
2581 and a conventional unicast DNS server on the same machine. If a
2582 conventional unicast DNS server wishes to implement mDNS as well,
2583 it can still do that, by opening two sockets. Having two different
2584 port numbers is important to allow this flexibility.
2586 * Some VPN software hijacks all outgoing traffic to port 53 and
2587 redirects it to a special DNS server set up to serve those VPN
2588 clients while they are connected to the corporate network. It is
2589 questionable whether this is the right thing to do, but it is
2590 common, and redirecting link-local multicast DNS packets to a
2591 remote server rarely produces any useful results. It does mean,
2592 for example, that the user becomes unable to access their local
2593 network printer sitting on their desk right next to their computer.
2594 Using a different UDP port eliminates this particular problem.
2596 * On many operating systems, unprivileged clients may not send or
2597 receive packets on low-numbered ports. This means that any client
2598 sending or receiving mDNS packets on port 53 would have to run
2599 as "root", which is an undesirable security risk. Using a higher-
2600 numbered UDP port eliminates this particular problem.
2610 Expires 10th February 2007 Cheshire & Krochmal [Page 45]
2612 Internet Draft Multicast DNS 10th August 2006
2615 22. Summary of Differences Between Multicast DNS and Unicast DNS
2617 The value of Multicast DNS is that it shares, as much as possible,
2618 the familiar APIs, naming syntax, resource record types, etc., of
2619 Unicast DNS. There are of course necessary differences by virtue of
2620 it using Multicast, and by virtue of it operating in a community of
2621 cooperating peers, rather than a precisely defined authoritarian
2622 hierarchy controlled by a strict chain of formal delegations from the
2623 top. These differences are listed below:
2627 * uses UDP port 5353 instead of port 53
2628 * operates in well-defined parts of the DNS namespace
2629 * uses UTF-8, and only UTF-8, to encode resource record names
2630 * defines a clear limit on the maximum legal domain name (255 bytes)
2631 * allows larger UDP packets
2632 * allows more than one question in a query packet
2633 * uses the Answer Section of a query to list Known Answers
2634 * uses the TC bit in a query to indicate additional Known Answers
2635 * uses the Authority Section of a query for probe tie-breaking
2636 * ignores the Query ID field (except for generating legacy responses)
2637 * doesn't require the question to be repeated in the response packet
2638 * uses gratuitous responses to announce new records to the peer group
2639 * defines a "unicast response" bit in the rrclass of query questions
2640 * defines a "cache flush" bit in the rrclass of response answers
2641 * uses DNS TTL 0 to indicate that a record has been deleted
2642 * monitors queries to perform Duplicate Question Suppression
2643 * monitors responses to perform Duplicate Answer Suppression...
2644 * ... and Ongoing Conflict Detection
2645 * ... and Opportunistic Caching
2668 Expires 10th February 2007 Cheshire & Krochmal [Page 46]
2670 Internet Draft Multicast DNS 10th August 2006
2672 23. Benefits of Multicast Responses
2674 Some people have argued that sending responses via multicast is
2675 inefficient on the network. In fact using multicast responses results
2676 in a net lowering of overall multicast traffic, for a variety of
2677 reasons, in addition to other benefits.
2679 * One multicast response can update the cache on all machines on the
2680 network. If another machine later wants to issue the same query, it
2681 already has the answer in its cache, so it may not need to even
2682 transmit that multicast query on the network at all.
2684 * When more than one machine has the same ongoing long-lived query
2685 running, every machine does not have to transmit its own
2686 independent query. When one machine transmits a query, all the
2687 other hosts see the answers, so they can suppress their own
2690 * When a host sees a multicast query, but does not see the corres-
2691 ponding multicast response, it can use this information to promptly
2692 delete stale data from its cache. To achieve the same level of
2693 user-interface quality and responsiveness without multicast
2694 responses would require lower cache lifetimes and more frequent
2695 network polling, resulting in a significantly higher packet rate.
2697 * Multicast responses allow passive conflict detection. Without this
2698 ability, some other conflict detection mechanism would be needed,
2699 imposing its own additional burden on the network.
2701 * When using delayed responses to reduce network collisions, clients
2702 need to maintain a list recording to whom each answer should be
2703 sent. The option of multicast responses allows clients with limited
2704 storage, which cannot store an arbitrarily long list of response
2705 addresses, to choose to fail-over to a single multicast response in
2706 place of multiple unicast responses, when appropriate.
2708 * In the case of overlayed subnets, multicast responses allow a
2709 receiver to know with certainty that a response originated on the
2710 local link, even when its source address may apparently suggest
2713 * Link-local multicast transcends virtually every conceivable network
2714 misconfiguration. Even if you have a collection of devices where
2715 every device's IP address, subnet mask, default gateway, and DNS
2716 server address are all wrong, packets sent by any of those devices
2717 addressed to a link-local multicast destination address will still
2718 be delivered to all peers on the local link. This can be extremely
2719 helpful when diagnosing and rectifying network problems, since
2720 it facilitates a direct communication channel between client and
2721 server that works without reliance on ARP, IP routing tables, etc.
2722 Being able to discover what IP address a device has (or thinks it
2723 has) is frequently a very valuable first step in diagnosing why it
2724 is unable to communicate on the local network.
2726 Expires 10th February 2007 Cheshire & Krochmal [Page 47]
2728 Internet Draft Multicast DNS 10th August 2006
2731 24. IPv6 Considerations
2733 An IPv4-only host and an IPv6-only host behave as "ships that pass in
2734 the night". Even if they are on the same Ethernet, neither is aware
2735 of the other's traffic. For this reason, each physical link may have
2736 *two* unrelated ".local." zones, one for IPv4 and one for IPv6.
2737 Since for practical purposes, a group of IPv4-only hosts and a group
2738 of IPv6-only hosts on the same Ethernet act as if they were on two
2739 entirely separate Ethernet segments, it is unsurprising that their
2740 use of the ".local." zone should occur exactly as it would if
2741 they really were on two entirely separate Ethernet segments.
2743 A dual-stack (v4/v6) host can participate in both ".local."
2744 zones, and should register its name(s) and perform its lookups both
2745 using IPv4 and IPv6. This enables it to reach, and be reached by,
2746 both IPv4-only and IPv6-only hosts. In effect this acts like a
2747 multi-homed host, with one connection to the logical "IPv4 Ethernet
2748 segment", and a connection to the logical "IPv6 Ethernet segment".
2751 24.1 IPv6 Multicast Addresses by Hashing
2753 Some discovery protocols use a range of multicast addresses, and
2754 determine the address to be used by a hash function of the name being
2755 sought. Queries are sent via multicast to the address as indicated by
2756 the hash function, and responses are returned to the querier via
2757 unicast. Particularly in IPv6, where multicast addresses are
2758 extremely plentiful, this approach is frequently advocated.
2760 There are some problems with this:
2762 * When a host has a large number of records with different names, the
2763 host may have to join a large number of multicast groups. This can
2764 place undue burden on the Ethernet hardware, which typically
2765 supports a limited number of multicast addresses efficiently. When
2766 this number is exceeded, the Ethernet hardware may have to resort
2767 to receiving all multicasts and passing them up to the host
2768 software for filtering, thereby defeating the point of using a
2769 multicast address range in the first place.
2771 * Multiple questions cannot be placed in one packet if they don't all
2772 hash to the same multicast address.
2774 * Duplicate Question Suppression doesn't work if queriers are not
2775 seeing each other's queries.
2777 * Duplicate Answer Suppression doesn't work if responders are not
2778 seeing each other's responses.
2780 * Opportunistic Caching doesn't work.
2782 * Ongoing Conflict Detection doesn't work.
2784 Expires 10th February 2007 Cheshire & Krochmal [Page 48]
2786 Internet Draft Multicast DNS 10th August 2006
2789 25. Security Considerations
2791 The algorithm for detecting and resolving name conflicts is, by its
2792 very nature, an algorithm that assumes cooperating participants. Its
2793 purpose is to allow a group of hosts to arrive at a mutually disjoint
2794 set of host names and other DNS resource record names, in the absence
2795 of any central authority to coordinate this or mediate disputes. In
2796 the absence of any higher authority to resolve disputes, the only
2797 alternative is that the participants must work together cooperatively
2798 to arrive at a resolution.
2800 In an environment where the participants are mutually antagonistic
2801 and unwilling to cooperate, other mechanisms are appropriate, like
2802 manually administered DNS.
2804 In an environment where there is a group of cooperating participants,
2805 but there may be other antagonistic participants on the same physical
2806 link, the cooperating participants need to use IPSEC signatures
2807 and/or DNSSEC [RFC 2535] signatures so that they can distinguish mDNS
2808 messages from trusted participants (which they process as usual) from
2809 mDNS messages from untrusted participants (which they silently
2812 When DNS queries for *global* DNS names are sent to the mDNS
2813 multicast address (during network outages which disrupt communication
2814 with the greater Internet) it is *especially* important to use
2815 DNSSEC, because the user may have the impression that he or she is
2816 communicating with some authentic host, when in fact he or she is
2817 really communicating with some local host that is merely masquerading
2818 as that name. This is less critical for names ending with ".local.",
2819 because the user should be aware that those names have only local
2820 significance and no global authority is implied.
2822 Most computer users neglect to type the trailing dot at the end of a
2823 fully qualified domain name, making it a relative domain name (e.g.
2824 "www.example.com"). In the event of network outage, attempts to
2825 positively resolve the name as entered will fail, resulting in
2826 application of the search list, including ".local.", if present.
2827 A malicious host could masquerade as "www.example.com" by answering
2828 the resulting Multicast DNS query for "www.example.com.local."
2829 To avoid this, a host MUST NOT append the search suffix
2830 ".local.", if present, to any relative (partially qualified)
2831 host name containing two or more labels. Appending ".local." to
2832 single-label relative host names is acceptable, since the user
2833 should have no expectation that a single-label host name will
2842 Expires 10th February 2007 Cheshire & Krochmal [Page 49]
2844 Internet Draft Multicast DNS 10th August 2006
2847 26. IANA Considerations
2849 IANA has allocated the IPv4 link-local multicast address 224.0.0.251
2850 for the use described in this document.
2852 IANA has allocated the IPv6 multicast address set FF0X::FB for the
2853 use described in this document. Only address FF02::FB (Link-Local
2854 Scope) is currently in use by deployed software, but it is possible
2855 that in future implementers may experiment with Multicast DNS using
2856 larger-scoped addresses, such as FF05::FB (Site-Local Scope).
2858 When this document is published, IANA should designate a list of
2859 domains which are deemed to have only link-local significance, as
2860 described in Section 12 of this document ("Special Characteristics of
2861 Multicast DNS Domains").
2863 The re-use of the top bit of the rrclass field in the Question and
2864 Answer Sections means that Multicast DNS can only carry DNS records
2865 with classes in the range 0-32767. Classes in the range 32768 to
2866 65535 are incompatible with Multicast DNS. However, since to-date
2867 only three DNS classes have been assigned by IANA (1, 3 and 4),
2868 and only one (1, "Internet") is actually in widespread use, this
2869 limitation is likely to remain a purely theoretical one.
2871 No other IANA services are required by this document.
2876 The concepts described in this document have been explored, developed
2877 and implemented with help from Freek Dijkstra, Erik Guttman, Paul
2878 Vixie, Bill Woodcock, and others.
2880 Special thanks go to Bob Bradley, Josh Graessley, Scott Herscher,
2881 Roger Pantos and Kiren Sekar for their significant contributions.
2884 28. Deployment History
2886 Multicast DNS client software first became available to the public
2887 in Mac OS 9 in 2001. Multicast DNS Responder software first began
2888 shipping to end users in large volumes (i.e. millions) with the
2889 launch of Mac OS X 10.2 Jaguar in August 2002, and became available
2890 for Microsoft Windows users with the launch of Apple's "Rendezvous
2891 for Windows" (now "Bonjour for Windows") in June 2004.
2893 Apple released the source code for the mDNSResponder daemon as Open
2894 Source in September 2002, first under Apple's standard Apple Public
2895 Source License, and then later, in August 2006, under the Apache
2896 License, Version 2.0.
2900 Expires 10th February 2007 Cheshire & Krochmal [Page 50]
2902 Internet Draft Multicast DNS 10th August 2006
2905 In addition to desktop and laptop computers running Mac OS X and
2906 Microsoft Windows, Multicast DNS is implemented in a wide range of
2907 hardware devices, such as Apple's "AirPort Extreme" and "AirPort
2908 Express" wireless base stations, home gateways from other vendors,
2909 network printers, network cameras, TiVo DVRs, etc.
2911 The Open Source community has produced many independent
2912 implementations of Multicast DNS, some in C like Apple's
2913 mDNSResponder daemon, and others in a variety of different languages
2914 including Java, Python, Perl, and C#/Mono.
2917 29. Copyright Notice
2919 Copyright (C) The Internet Society (2006).
2921 This document is subject to the rights, licenses and restrictions
2922 contained in BCP 78, and except as set forth therein, the authors
2923 retain all their rights. For the purposes of this document,
2924 the term "BCP 78" refers exclusively to RFC 3978, "IETF Rights
2925 in Contributions", published March 2005.
2927 This document and the information contained herein are provided on an
2928 "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
2929 OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
2930 ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED,
2931 INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE
2932 INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED
2933 WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
2936 30. Normative References
2938 [RFC 1034] Mockapetris, P., "Domain Names - Concepts and
2939 Facilities", STD 13, RFC 1034, November 1987.
2941 [RFC 1035] Mockapetris, P., "Domain Names - Implementation and
2942 Specifications", STD 13, RFC 1035, November 1987.
2944 [RFC 2119] Bradner, S., "Key words for use in RFCs to Indicate
2945 Requirement Levels", RFC 2119, March 1997.
2947 [RFC 3629] Yergeau, F., "UTF-8, a transformation format of ISO
2948 10646", RFC 3629, November 2003.
2950 [UAX15] "Unicode Normalization Forms"
2951 http://www.unicode.org/reports/tr15/
2958 Expires 10th February 2007 Cheshire & Krochmal [Page 51]
2960 Internet Draft Multicast DNS 10th August 2006
2963 31. Informative References
2965 [dotlocal] <http://www.dotlocal.org/>
2967 [djbdl] <http://cr.yp.to/djbdns/dot-local.html>
2969 [DNS-SD] Cheshire, S., and M. Krochmal, "DNS-Based Service
2970 Discovery", Internet-Draft (work in progress),
2971 draft-cheshire-dnsext-dns-sd-04.txt, August 2006.
2973 [IEEE802] IEEE Standards for Local and Metropolitan Area Networks:
2974 Overview and Architecture.
2975 Institute of Electrical and Electronic Engineers,
2976 IEEE Standard 802, 1990.
2978 [NBP] Cheshire, S., and M. Krochmal,
2979 "Requirements for a Protocol to Replace AppleTalk NBP",
2980 Internet-Draft (work in progress),
2981 draft-cheshire-dnsext-nbp-05.txt, August 2006.
2983 [RFC 2136] Vixie, P., et al., "Dynamic Updates in the Domain Name
2984 System (DNS UPDATE)", RFC 2136, April 1997.
2986 [RFC 2462] S. Thomson and T. Narten, "IPv6 Stateless Address
2987 Autoconfiguration", RFC 2462, December 1998.
2989 [RFC 2535] Eastlake, D., "Domain Name System Security Extensions",
2990 RFC 2535, March 1999.
2992 [RFC 2606] Eastlake, D., and A. Panitz, "Reserved Top Level DNS
2993 Names", RFC 2606, June 1999.
2995 [RFC 2860] Carpenter, B., Baker, F. and M. Roberts, "Memorandum
2996 of Understanding Concerning the Technical Work of the
2997 Internet Assigned Numbers Authority", RFC 2860, June
3000 [RFC 3492] Costello, A., "Punycode: A Bootstring encoding of
3001 Unicode for use with Internationalized Domain Names
3002 in Applications (IDNA)", RFC 3492, March 2003.
3004 [RFC 3927] Cheshire, S., B. Aboba, and E. Guttman,
3005 "Dynamic Configuration of IPv4 Link-Local Addresses",
3008 [ZC] Williams, A., "Requirements for Automatic Configuration
3009 of IP Hosts", Internet-Draft (work in progress),
3010 draft-ietf-zeroconf-reqts-12.txt, September 2002.
3016 Expires 10th February 2007 Cheshire & Krochmal [Page 52]
3018 Internet Draft Multicast DNS 10th August 2006
3021 32. Authors' Addresses
3024 Apple Computer, Inc.
3030 Phone: +1 408 974 3207
3031 EMail: rfc [at] stuartcheshire [dot] org
3035 Apple Computer, Inc.
3041 Phone: +1 408 974 4368
3042 EMail: marc [at] apple [dot] com
3074 Expires 10th February 2007 Cheshire & Krochmal [Page 53]