Network Working Group                                     A. Gulbrandsen
Request for Comments: 2782                            Troll Technologies
Obsoletes: 2052                                                 P. Vixie
Category: Standards Track                   Internet Software Consortium
                                                              L. Esibov
                                                        Microsoft Corp.
                                                          February 2000


      A DNS RR for specifying the location of services (DNS SRV)

Status of this Memo

  This document specifies an Internet standards track protocol for the
  Internet community, and requests discussion and suggestions for
  improvements.  Please refer to the current edition of the "Internet
  Official Protocol Standards" (STD 1) for the standardization state
  and status of this protocol.  Distribution of this memo is unlimited.

Copyright Notice

  Copyright (C) The Internet Society (2000).  All Rights Reserved.

Abstract

  This document describes a DNS RR which specifies the location of the
  server(s) for a specific protocol and domain.

Overview and rationale

  Currently, one must either know the exact address of a server to
  contact it, or broadcast a question.

  The SRV RR allows administrators to use several servers for a single
  domain, to move services from host to host with little fuss, and to
  designate some hosts as primary servers for a service and others as
  backups.

  Clients ask for a specific service/protocol for a specific domain
  (the word domain is used here in the strict RFC 1034 sense), and get
  back the names of any available servers.

  Note that where this document refers to "address records", it means A
  RR's, AAAA RR's, or their most modern equivalent.







Gulbrandsen, et al.         Standards Track                     [Page 1]

RFC 2782                       DNS SRV RR                  February 2000


Definitions

  The key words "MUST", "MUST NOT", "SHOULD", "SHOULD NOT" and "MAY"
  used in this document are to be interpreted as specified in [BCP 14].
  Other terms used in this document are defined in the DNS
  specification, RFC 1034.

Applicability Statement

  In general, it is expected that SRV records will be used by clients
  for applications where the relevant protocol specification indicates
  that clients should use the SRV record. Such specification MUST
  define the symbolic name to be used in the Service field of the SRV
  record as described below. It also MUST include security
  considerations. Service SRV records SHOULD NOT be used in the absence
  of such specification.

Introductory example

  If a SRV-cognizant LDAP client wants to discover a LDAP server that
  supports TCP protocol and provides LDAP service for the domain
  example.com., it does a lookup of

     _ldap._tcp.example.com

  as described in [ARM].  The example zone file near the end of this
  memo contains answering RRs for an SRV query.

  Note: LDAP is chosen as an example for illustrative purposes only,
  and the LDAP examples used in this document should not be considered
  a definitive statement on the recommended way for LDAP to use SRV
  records. As described in the earlier applicability section, consult
  the appropriate LDAP documents for the recommended procedures.

The format of the SRV RR

  Here is the format of the SRV RR, whose DNS type code is 33:

       _Service._Proto.Name TTL Class SRV Priority Weight Port Target

       (There is an example near the end of this document.)

  Service
       The symbolic name of the desired service, as defined in Assigned
       Numbers [STD 2] or locally.  An underscore (_) is prepended to
       the service identifier to avoid collisions with DNS labels that
       occur in nature.




Gulbrandsen, et al.         Standards Track                     [Page 2]

RFC 2782                       DNS SRV RR                  February 2000


       Some widely used services, notably POP, don't have a single
       universal name.  If Assigned Numbers names the service
       indicated, that name is the only name which is legal for SRV
       lookups.  The Service is case insensitive.

  Proto
       The symbolic name of the desired protocol, with an underscore
       (_) prepended to prevent collisions with DNS labels that occur
       in nature.  _TCP and _UDP are at present the most useful values
       for this field, though any name defined by Assigned Numbers or
       locally may be used (as for Service).  The Proto is case
       insensitive.

  Name
       The domain this RR refers to.  The SRV RR is unique in that the
       name one searches for is not this name; the example near the end
       shows this clearly.

  TTL
       Standard DNS meaning [RFC 1035].

  Class
       Standard DNS meaning [RFC 1035].   SRV records occur in the IN
       Class.

  Priority
       The priority of this target host.  A client MUST attempt to
       contact the target host with the lowest-numbered priority it can
       reach; target hosts with the same priority SHOULD be tried in an
       order defined by the weight field.  The range is 0-65535.  This
       is a 16 bit unsigned integer in network byte order.

  Weight
       A server selection mechanism.  The weight field specifies a
       relative weight for entries with the same priority. Larger
       weights SHOULD be given a proportionately higher probability of
       being selected. The range of this number is 0-65535.  This is a
       16 bit unsigned integer in network byte order.  Domain
       administrators SHOULD use Weight 0 when there isn't any server
       selection to do, to make the RR easier to read for humans (less
       noisy).  In the presence of records containing weights greater
       than 0, records with weight 0 should have a very small chance of
       being selected.

       In the absence of a protocol whose specification calls for the
       use of other weighting information, a client arranges the SRV
       RRs of the same Priority in the order in which target hosts,




Gulbrandsen, et al.         Standards Track                     [Page 3]

RFC 2782                       DNS SRV RR                  February 2000


       specified by the SRV RRs, will be contacted. The following
       algorithm SHOULD be used to order the SRV RRs of the same
       priority:

       To select a target to be contacted next, arrange all SRV RRs
       (that have not been ordered yet) in any order, except that all
       those with weight 0 are placed at the beginning of the list.

       Compute the sum of the weights of those RRs, and with each RR
       associate the running sum in the selected order. Then choose a
       uniform random number between 0 and the sum computed
       (inclusive), and select the RR whose running sum value is the
       first in the selected order which is greater than or equal to
       the random number selected. The target host specified in the
       selected SRV RR is the next one to be contacted by the client.
       Remove this SRV RR from the set of the unordered SRV RRs and
       apply the described algorithm to the unordered SRV RRs to select
       the next target host.  Continue the ordering process until there
       are no unordered SRV RRs.  This process is repeated for each
       Priority.

  Port
       The port on this target host of this service.  The range is 0-
       65535.  This is a 16 bit unsigned integer in network byte order.
       This is often as specified in Assigned Numbers but need not be.

  Target
       The domain name of the target host.  There MUST be one or more
       address records for this name, the name MUST NOT be an alias (in
       the sense of RFC 1034 or RFC 2181).  Implementors are urged, but
       not required, to return the address record(s) in the Additional
       Data section.  Unless and until permitted by future standards
       action, name compression is not to be used for this field.

       A Target of "." means that the service is decidedly not
       available at this domain.

Domain administrator advice

  Expecting everyone to update their client applications when the first
  server publishes a SRV RR is futile (even if desirable).  Therefore
  SRV would have to coexist with address record lookups for existing
  protocols, and DNS administrators should try to provide address
  records to support old clients:

     - Where the services for a single domain are spread over several
       hosts, it seems advisable to have a list of address records at
       the same DNS node as the SRV RR, listing reasonable (if perhaps



Gulbrandsen, et al.         Standards Track                     [Page 4]

RFC 2782                       DNS SRV RR                  February 2000


       suboptimal) fallback hosts for Telnet, NNTP and other protocols
       likely to be used with this name.  Note that some programs only
       try the first address they get back from e.g. gethostbyname(),
       and we don't know how widespread this behavior is.

     - Where one service is provided by several hosts, one can either
       provide address records for all the hosts (in which case the
       round-robin mechanism, where available, will share the load
       equally) or just for one (presumably the fastest).

     - If a host is intended to provide a service only when the main
       server(s) is/are down, it probably shouldn't be listed in
       address records.

     - Hosts that are referenced by backup address records must use the
       port number specified in Assigned Numbers for the service.

     - Designers of future protocols for which "secondary servers" is
       not useful (or meaningful) may choose to not use SRV's support
       for secondary servers.  Clients for such protocols may use or
       ignore SRV RRs with Priority higher than the RR with the lowest
       Priority for a domain.

  Currently there's a practical limit of 512 bytes for DNS replies.
  Until all resolvers can handle larger responses, domain
  administrators are strongly advised to keep their SRV replies below
  512 bytes.

  All round numbers, wrote Dr. Johnson, are false, and these numbers
  are very round: A reply packet has a 30-byte overhead plus the name
  of the service ("_ldap._tcp.example.com" for instance); each SRV RR
  adds 20 bytes plus the name of the target host; each NS RR in the NS
  section is 15 bytes plus the name of the name server host; and
  finally each A RR in the additional data section is 20 bytes or so,
  and there are A's for each SRV and NS RR mentioned in the answer.
  This size estimate is extremely crude, but shouldn't underestimate
  the actual answer size by much.  If an answer may be close to the
  limit, using a DNS query tool (e.g. "dig") to look at the actual
  answer is a good idea.

The "Weight" field

  Weight, the server selection field, is not quite satisfactory, but
  the actual load on typical servers changes much too quickly to be
  kept around in DNS caches.  It seems to the authors that offering
  administrators a way to say "this machine is three times as fast as
  that one" is the best that can practically be done.




Gulbrandsen, et al.         Standards Track                     [Page 5]

RFC 2782                       DNS SRV RR                  February 2000


  The only way the authors can see of getting a "better" load figure is
  asking a separate server when the client selects a server and
  contacts it.  For short-lived services an extra step in the
  connection establishment seems too expensive, and for long-lived
  services, the load figure may well be thrown off a minute after the
  connection is established when someone else starts or finishes a
  heavy job.

  Note: There are currently various experiments at providing relative
  network proximity estimation, available bandwidth estimation, and
  similar services.  Use of the SRV record with such facilities, and in
  particular the interpretation of the Weight field when these
  facilities are used, is for further study.  Weight is only intended
  for static, not dynamic, server selection.  Using SRV weight for
  dynamic server selection would require assigning unreasonably short
  TTLs to the SRV RRs, which would limit the usefulness of the DNS
  caching mechanism, thus increasing overall network load and
  decreasing overall reliability.  Server selection via SRV is only
  intended to express static information such as "this server has a
  faster CPU than that one" or "this server has a much better network
  connection than that one".

The Port number

  Currently, the translation from service name to port number happens
  at the client, often using a file such as /etc/services.

  Moving this information to the DNS makes it less necessary to update
  these files on every single computer of the net every time a new
  service is added, and makes it possible to move standard services out
  of the "root-only" port range on unix.

Usage rules

  A SRV-cognizant client SHOULD use this procedure to locate a list of
  servers and connect to the preferred one:

       Do a lookup for QNAME=_service._protocol.target, QCLASS=IN,
       QTYPE=SRV.

       If the reply is NOERROR, ANCOUNT>0 and there is at least one
       SRV RR which specifies the requested Service and Protocol in
       the reply:

           If there is precisely one SRV RR, and its Target is "."
           (the root domain), abort.





Gulbrandsen, et al.         Standards Track                     [Page 6]

RFC 2782                       DNS SRV RR                  February 2000


           Else, for all such RR's, build a list of (Priority, Weight,
           Target) tuples

           Sort the list by priority (lowest number first)

           Create a new empty list

           For each distinct priority level
               While there are still elements left at this priority
               level

                   Select an element as specified above, in the
                   description of Weight in "The format of the SRV
                   RR" Section, and move it to the tail of the new
                   list

           For each element in the new list

               query the DNS for address records for the Target or
               use any such records found in the Additional Data
               section of the earlier SRV response.

               for each address record found, try to connect to the
              (protocol, address, service).

       else

           Do a lookup for QNAME=target, QCLASS=IN, QTYPE=A

           for each address record found, try to connect to the
          (protocol, address, service)

Notes:

  - Port numbers SHOULD NOT be used in place of the symbolic service
    or protocol names (for the same reason why variant names cannot
    be allowed: Applications would have to do two or more lookups).

  - If a truncated response comes back from an SRV query, the rules
    described in [RFC 2181] shall apply.

  - A client MUST parse all of the RR's in the reply.

  - If the Additional Data section doesn't contain address records
    for all the SRV RR's and the client may want to connect to the
    target host(s) involved, the client MUST look up the address
    record(s).  (This happens quite often when the address record
    has shorter TTL than the SRV or NS RR's.)



Gulbrandsen, et al.         Standards Track                     [Page 7]

RFC 2782                       DNS SRV RR                  February 2000


  - Future protocols could be designed to use SRV RR lookups as the
    means by which clients locate their servers.

Fictional example

  This example uses fictional service "foobar" as an aid in
  understanding SRV records. If ever service "foobar" is implemented,
  it is not intended that it will necessarily use SRV records.  This is
  (part of) the zone file for example.com, a still-unused domain:

     $ORIGIN example.com.
     @               SOA server.example.com. root.example.com. (
                         1995032001 3600 3600 604800 86400 )
                     NS  server.example.com.
                     NS  ns1.ip-provider.net.
                     NS  ns2.ip-provider.net.
     ; foobar - use old-slow-box or new-fast-box if either is
     ; available, make three quarters of the logins go to
     ; new-fast-box.
     _foobar._tcp    SRV 0 1 9 old-slow-box.example.com.
                      SRV 0 3 9 new-fast-box.example.com.
     ; if neither old-slow-box or new-fast-box is up, switch to
     ; using the sysdmin's box and the server
                      SRV 1 0 9 sysadmins-box.example.com.
                      SRV 1 0 9 server.example.com.
     server           A   172.30.79.10
     old-slow-box     A   172.30.79.11
     sysadmins-box    A   172.30.79.12
     new-fast-box     A   172.30.79.13
     ; NO other services are supported
     *._tcp          SRV  0 0 0 .
     *._udp          SRV  0 0 0 .



















Gulbrandsen, et al.         Standards Track                     [Page 8]

RFC 2782                       DNS SRV RR                  February 2000


  In this example, a client of the "foobar" service in the
  "example.com." domain needs an SRV lookup of
  "_foobar._tcp.example.com." and possibly A lookups of "new-fast-
  box.example.com." and/or the other hosts named.  The size of the SRV
  reply is approximately 365 bytes:

     30 bytes general overhead
     20 bytes for the query string, "_foobar._tcp.example.com."
     130 bytes for 4 SRV RR's, 20 bytes each plus the lengths of "new-
       fast-box", "old-slow-box", "server" and "sysadmins-box" -
       "example.com" in the query section is quoted here and doesn't
       need to be counted again.
     75 bytes for 3 NS RRs, 15 bytes each plus the lengths of "server",
       "ns1.ip-provider.net." and "ns2" - again, "ip-provider.net." is
       quoted and only needs to be counted once.
     120 bytes for the 6 address records (assuming IPv4 only) mentioned
       by the SRV and NS RR's.

IANA Considerations

  The IANA has assigned RR type value 33 to the SRV RR.  No other IANA
  services are required by this document.

Changes from RFC 2052

  This document obsoletes RFC 2052.   The major change from that
  previous, experimental, version of this specification is that now the
  protocol and service labels are prepended with an underscore, to
  lower the probability of an accidental clash with a similar name used
  for unrelated purposes.  Aside from that, changes are only intended
  to increase the clarity and completeness of the document. This
  document especially clarifies the use of the Weight field of the SRV
  records.

Security Considerations

  The authors believe this RR to not cause any new security problems.
  Some problems become more visible, though.

  - The ability to specify ports on a fine-grained basis obviously
    changes how a router can filter packets.  It becomes impossible
    to block internal clients from accessing specific external
    services, slightly harder to block internal users from running
    unauthorized services, and more important for the router
    operations and DNS operations personnel to cooperate.

  - There is no way a site can keep its hosts from being referenced
    as servers.  This could lead to denial of service.



Gulbrandsen, et al.         Standards Track                     [Page 9]

RFC 2782                       DNS SRV RR                  February 2000


  - With SRV, DNS spoofers can supply false port numbers, as well as
    host names and addresses.   Because this vulnerability exists
    already, with names and addresses, this is not a new
    vulnerability, merely a slightly extended one, with little
    practical effect.

References

  STD 2:    Reynolds, J., and J. Postel, "Assigned Numbers", STD 2, RFC
            1700, October 1994.

  RFC 1034: Mockapetris, P., "Domain names - concepts and facilities",
            STD 13, RFC 1034, November 1987.

  RFC 1035: Mockapetris, P., "Domain names - Implementation and
            Specification", STD 13, RFC 1035, November 1987.

  RFC 974:  Partridge, C., "Mail routing and the domain system", STD
            14, RFC 974, January 1986.

  BCP 14:   Bradner, S., "Key words for use in RFCs to Indicate
            Requirement Levels", BCP 14, RFC 2119, March 1997.

  RFC 2181: Elz, R. and R. Bush, "Clarifications to the DNS
            Specification", RFC 2181, July 1997.

  RFC 2219: Hamilton, M. and R. Wright, "Use of DNS Aliases for Network
            Services", BCP 17, RFC 2219, October 1997.

  BCP 14:   Bradner, S., "Key words for use in RFCs to Indicate
            Requirement Levels", BCP 14, RFC 2119, March 1997.

  ARM:      Armijo, M., Esibov, L. and P. Leach, "Discovering LDAP
            Services with DNS", Work in Progress.

  KDC-DNS:  Hornstein, K. and J. Altman, "Distributing Kerberos KDC and
            Realm Information with DNS", Work in Progress.














Gulbrandsen, et al.         Standards Track                    [Page 10]

RFC 2782                       DNS SRV RR                  February 2000


Acknowledgements

  The algorithm used to select from the weighted SRV RRs of equal
  priority is adapted from one supplied by Dan Bernstein.

Authors' Addresses

  Arnt Gulbrandsen
  Troll Tech
  Waldemar Thranes gate 98B
  N-0175 Oslo, Norway

  Fax:   +47 22806380
  Phone: +47 22806390
  EMail: [email protected]


  Paul Vixie
  Internet Software Consortium
  950 Charter Street
  Redwood City, CA 94063

  Phone: +1 650 779 7001


  Levon Esibov
  Microsoft Corporation
  One Microsoft Way
  Redmond, WA 98052

  EMail: [email protected]




















Gulbrandsen, et al.         Standards Track                    [Page 11]

RFC 2782                       DNS SRV RR                  February 2000


Full Copyright Statement

  Copyright (C) The Internet Society (2000).  All Rights Reserved.

  This document and translations of it may be copied and furnished to
  others, and derivative works that comment on or otherwise explain it
  or assist in its implementation may be prepared, copied, published
  and distributed, in whole or in part, without restriction of any
  kind, provided that the above copyright notice and this paragraph are
  included on all such copies and derivative works.  However, this
  document itself may not be modified in any way, such as by removing
  the copyright notice or references to the Internet Society or other
  Internet organizations, except as needed for the purpose of
  developing Internet standards in which case the procedures for
  copyrights defined in the Internet Standards process must be
  followed, or as required to translate it into languages other than
  English.

  The limited permissions granted above are perpetual and will not be
  revoked by the Internet Society or its successors or assigns.

  This document and the information contained herein is provided on an
  "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
  TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING
  BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION
  HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF
  MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

Acknowledgement

  Funding for the RFC Editor function is currently provided by the
  Internet Society.



















Gulbrandsen, et al.         Standards Track                    [Page 12]