<pre>Network Working Group                                                IAB
Request for Comments: 5507                             P. Faltstrom, Ed.
Category: Informational                                  R. Austein, Ed.
                                                            P. Koch, Ed.
                                                              April 2009


                 <span class="h1">Design Choices When Expanding the DNS</span>

Status of This Memo

   This memo provides information for the Internet community.  It does
   not specify an Internet standard of any kind.  Distribution of this
   memo is unlimited.

Copyright Notice

   Copyright (c) 2009 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to <a href="https://www.rfc-editor.org/bcp/bcp78">BCP 78</a> and the IETF Trust's Legal
   Provisions Relating to IETF Documents in effect on the date of
   publication of this document (<a href="http://trustee.ietf.org/license-info">http://trustee.ietf.org/license-info</a>).
   Please review these documents carefully, as they describe your rights
   and restrictions with respect to this document.


Abstract

   This note discusses how to extend the DNS with new data for a new
   application.  DNS extension discussions too often focus on reuse of
   the TXT Resource Record Type.  This document lists different
   mechanisms to extend the DNS, and concludes that the use of a new DNS
   Resource Record Type is the best solution.

















<span class="grey">IAB, et al.                  Informational                      [Page 1]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-2" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


Table of Contents

   <a href="#section-1">1</a>. Introduction ....................................................<a href="#page-3">3</a>
   <a href="#section-2">2</a>. Background ......................................................<a href="#page-4">4</a>
   <a href="#section-3">3</a>. Extension Mechanisms ............................................<a href="#page-5">5</a>
      3.1. Place Selectors inside the RDATA of Existing
           Resource Record Types ......................................<a href="#page-5">5</a>
      <a href="#section-3.2">3.2</a>. Add a Prefix to the Owner Name .............................<a href="#page-6">6</a>
      <a href="#section-3.3">3.3</a>. Add a Suffix to the Owner Name .............................<a href="#page-7">7</a>
      <a href="#section-3.4">3.4</a>. Add a New Class ............................................<a href="#page-8">8</a>
      <a href="#section-3.5">3.5</a>. Add a New Resource Record Type .............................<a href="#page-8">8</a>
   <a href="#section-4">4</a>. Zone Boundaries are Invisible to Applications ...................<a href="#page-9">9</a>
   5. Why Adding a New Resource Record Type Is the Preferred
      Solution .......................................................<a href="#page-10">10</a>
   <a href="#section-6">6</a>. Conclusion and Recommendation ..................................<a href="#page-14">14</a>
   <a href="#section-7">7</a>. Creating a New Resource Record Type ............................<a href="#page-14">14</a>
   <a href="#section-8">8</a>. Security Considerations ........................................<a href="#page-15">15</a>
   <a href="#section-9">9</a>. Acknowledgements ...............................................<a href="#page-15">15</a>
   <a href="#section-10">10</a>. IAB Members at the Time of This Writing .......................<a href="#page-16">16</a>
   <a href="#section-11">11</a>. References ....................................................<a href="#page-16">16</a>
      <a href="#section-11.1">11.1</a>. Normative References .....................................<a href="#page-16">16</a>
      <a href="#section-11.2">11.2</a>. Informative References ...................................<a href="#page-16">16</a>





























<span class="grey">IAB, et al.                  Informational                      [Page 2]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-3" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


<span class="h2"><a class="selflink" id="section-1" href="#section-1">1</a>.  Introduction</span>

   The DNS stores multiple categories of data.  The two most commonly
   used categories are infrastructure data for the DNS system itself (NS
   and SOA Resource Records) and data that have to do with mappings
   between domain names and IP addresses (A, AAAA, and PTR Resource
   Records).  There are other categories as well, some of which are tied
   to specific applications like email (MX Resource Records), while
   others are generic Resource Record Types used to convey information
   for multiple protocols (SRV and NAPTR Resource Records).

   When storing data in the DNS for a new application, the goal must be
   to store data in such a way that the application can query for the
   data it wants, while minimizing both the impact on existing
   applications and the amount of extra data transferred to the client.
   This implies that a number of design choices have to be made, where
   the most important is to ensure that a precise selection of what data
   to return must be made already in the query.  A query consists of a
   triple: {Owner (or name), Resource Record Class, Resource Record
   Type}.

   Historically, extending the DNS to store application data tied to a
   domain name has been done in different ways at different times.  MX
   Resource Records were created as a new Resource Record Type
   specifically designed to support electronic mail.  SRV records are a
   generic type that use a prefixing scheme in combination with a base
   domain name.  NAPTR records add selection data inside the RDATA.  It
   is clear that the methods used to add new data types to the DNS have
   been inconsistent, and the purpose of this document is to attempt to
   clarify the implications of each of these methods, both for the
   applications that use them and for the rest of the DNS.

   This document talks extensively about use of DNS wildcards.  Many
   people might think use of wildcards is not something that happens
   today.  In reality though, wildcards are in use, especially for
   certain application-specific data such as MX Resource Records.
   Because of this, the choice has to be made with the existence of
   wildcards in mind.

   Another overall issue that must be taken into account is what the new
   data in the DNS are to describe.  In some cases, they might be
   completely new data.  In other cases, they might be metadata tied to
   data that already exist in the DNS.  Examples of new data are key
   information for the Secure SHell (SSH) Protocol and data used for
   authenticating the sender of email messages (metadata tied to MX
   Resource Records).  If the new data are tied to data that already
   exist in the DNS, an analysis should be made as to whether having
   (for example) address records and SSH key information in different



<span class="grey">IAB, et al.                  Informational                      [Page 3]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-4" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


   DNS zones is a problem or if it is a bonus, and if it is a problem,
   whether the specification must require all of the related data to be
   in the same zone.  One specific difference between having the records
   in the same zone or not has to do with maintenance of the records.
   If they are in the same zone, the same maintainer (from a DNS
   perspective) manages the two records.  Specifically, they must be
   signed with the same DNSSEC keys if DNSSEC is in use.

   This document does not talk about what one should store in the DNS.
   It also doesn't discuss whether the DNS should be used for service
   discovery, or whether the DNS should be used for storage of data
   specific to the service.  In general, the DNS is a protocol that,
   apart from holding metadata that makes the DNS itself function (NS,
   SOA, DNSSEC Resource Record Types, etc.), only holds references to
   service locations (SRV, NAPTR, A, AAAA Resource Record Types) --
   though there are exceptions, such as MX Resource Records.

<span class="h2"><a class="selflink" id="section-2" href="#section-2">2</a>.  Background</span>

   See <a href="./rfc5395">RFC 5395</a> [<a href="./rfc5395" title="&quot;Domain Name System (DNS) IANA Considerations&quot;">RFC5395</a>] for a brief summary of the DNS query
   structure.  Readers interested in the full story should start with
   the base DNS specification in <a href="./rfc1035">RFC 1035</a> [<a href="./rfc1035" title="&quot;Domain names - implementation and specification&quot;">RFC1035</a>] and continue with
   the various documents that update, clarify, and extend the base
   specification.

   When composing a DNS query, the parameters used by the protocol are a
   {owner, class, type} triple.  Every Resource Record matching such a
   triple is said to belong to the same Resource Record Set (RRSet), and
   the whole RRSet is always returned to the client that queries for it.
   Splitting an RRSet is a protocol violation (sending a partial RRSet,
   not truncating the DNS response), because it can result in coherency
   problems with the DNS caching mechanism.  See <a href="./rfc2181#section-5">Section&nbsp;5 of [RFC2181]</a>
   for more information.

   Some discussions around extensions to the DNS include arguments
   around MTU size.  Note that most discussions about DNS and MTU size
   are about the size of the whole DNS packet, not about the size of a
   single RRSet.

   Almost all DNS query traffic is carried over UDP, where a DNS message
   must fit within a single UDP packet.  DNS response messages are
   almost always larger than DNS query messages, so message size issues
   are almost always about responses, not queries.  The base DNS
   specification limits DNS messages over UDP to 512 octets; EDNS0
   [<a href="./rfc2671" title="&quot;Extension Mechanisms for DNS (EDNS0)&quot;">RFC2671</a>] specifies a mechanism by which a client can signal its
   willingness to receive larger responses, but deployment of EDNS0 is
   not universal, in part because of firewalls that block fragmented UDP
   packets or EDNS0.  If a response message won't fit in a single



<span class="grey">IAB, et al.                  Informational                      [Page 4]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-5" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


   packet, the name server returns a truncated response, at which point
   the client may retry using TCP.  DNS queries over TCP are not subject
   to this length limitation, but TCP imposes significantly higher per-
   query overhead on name servers than UDP.  It is also the case that
   the policies in deployed firewalls far too often are such that they
   block DNS over TCP, so using TCP might not in reality be an option.
   There are also risks (although possibly small) that a change of
   routing while a TCP flow is open creates problems when the DNS
   servers are deployed in an anycast environment.

<span class="h2"><a class="selflink" id="section-3" href="#section-3">3</a>.  Extension Mechanisms</span>

   The DNS protocol is intended to be extensible to support new kinds of
   data.  This section examines the various ways in which this sort of
   extension can be accomplished.

<span class="h3"><a class="selflink" id="section-3.1" href="#section-3.1">3.1</a>.  Place Selectors inside the RDATA of Existing Resource Record Types</span>

   For a given query name, one might choose to have a single RRSet (all
   Resource Records sharing the same {owner, class, type} triple) shared
   by multiple applications, and have the different applications use
   selectors within the Resource Record data (RDATA) to determine which
   records are intended for which applications.  This sort of selector
   mechanism is usually referred to "subtyping", because it is in effect
   creating an additional type subsystem within a single DNS Resource
   Record Type.

   Examples of subtyping include NAPTR Resource Records [<a href="./rfc3761" title="&quot;The E.164 to Uniform Resource Identifiers (URI) Dynamic Delegation Discovery System (DDDS) Application (ENUM)&quot;">RFC3761</a>] and
   the original DNSSEC KEY Resource Record Type [<a href="./rfc2535" title="&quot;Domain Name System Security Extensions&quot;">RFC2535</a>] (which was
   later updated by <a href="./rfc3445">RFC 3445</a> [<a href="./rfc3445" title="&quot;Limiting the Scope of the KEY Resource Record (RR)&quot;">RFC3445</a>], and obsoleted by <a href="./rfc4033">RFC 4033</a>
   [<a href="./rfc4033" title="&quot;DNS Security Introduction and Requirements&quot;">RFC4033</a>], <a href="./rfc4034">RFC 4034</a> [<a href="./rfc4034" title="&quot;Resource Records for the DNS Security Extensions&quot;">RFC4034</a>] and <a href="./rfc4035">RFC 4035</a> [<a href="./rfc4035" title="&quot;Protocol Modifications for the DNS Security Extensions&quot;">RFC4035</a>]).

   All DNS subtyping schemes share a common weakness: with subtyping
   schemes, it is impossible for a client to query for just the data it
   wants.  Instead, the client must fetch the entire RRSet, then select
   the Resource Records in which it is interested.  Furthermore, since
   DNSSEC signatures operate on complete RRSets, the entire RRSet must
   be re-signed if any Resource Record in it changes.  As a result, each
   application that uses a subtyped Resource Record incurs higher
   overhead than any of the applications would have incurred had they
   not been using a subtyping scheme.  The fact the RRSet is always
   passed around as an indivisible unit increases the risk the RRSet
   will not fit in a UDP packet, which in turn increases the risk that
   the client will have to retry the query with TCP, which substantially
   increases the load on the name server.  More precisely: having one
   query fail over to TCP is not a big deal, but since the typical ratio





<span class="grey">IAB, et al.                  Informational                      [Page 5]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-6" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


   of clients to servers in today's deployed DNS is very high, having a
   substantial number of DNS messages fail over to TCP may cause the
   queried name servers to be overloaded by TCP overhead.

   Because of the size limitations, using a subtyping scheme to list a
   large number of services for a single domain name risks triggering
   truncation and fallback to TCP, which may in turn force the zone
   administrator to announce only a subset of available services.

<span class="h3"><a class="selflink" id="section-3.2" href="#section-3.2">3.2</a>.  Add a Prefix to the Owner Name</span>

   By adding an application-specific prefix to a domain name, we get a
   different {owner, class, type} triple, and therefore a different
   RRSet.  One problem with adding prefixes has to do with wildcards,
   especially if one has records like:

   *.example.com. IN MX 1 mail.example.com.

   and one wants records tied to those names.  Suppose one creates the
   prefix "_mail".  One would then have to say something like:

   _mail.*.example.com. IN X-FOO A B C D

   but DNS wildcards only work with the "*" as the leftmost token in the
   domain name (see also <a href="./rfc4592">RFC 4592</a> [<a href="./rfc4592" title="&quot;The Role of Wildcards in the Domain Name System&quot;">RFC4592</a>]).

   There have been proposals to deal with the problem that DNS wildcards
   are always terminal records.  These proposals introduce an additional
   set of trade-offs that would need to be taken into account when
   assessing which extension mechanism to choose.  Aspects of extra
   response time needed to perform the extra queries, costs of pre-
   calculation of possible answers, or the costs induced to the system
   as a whole come to mind.  At the time of writing, none of these
   proposals has been published as Standards Track RFCs.

   Even when a specific prefix is chosen, the data will still have to be
   stored in some Resource Record Type.  This Resource Record Type can
   be either a new Resource Record Type or an existing Resource Record
   Type that has an appropriate format to store the data.  One also
   might need some other selection mechanism, such as the ability to
   distinguish between the records in an RRSet, given they have the same
   Resource Record Type.  Because of this, one needs to both register a
   unique prefix and define what Resource Record Type is to be used for
   this specific service.







<span class="grey">IAB, et al.                  Informational                      [Page 6]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-7" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


   If the record has some relationship with another record in the zone,
   the fact that the two records can be in different zones might have
   implications on the trust the application has in the records.  For
   example:

   example.com.      IN MX    10 mail.example.com.
   _foo.example.com. IN X-BAR "metadata for the mail service"

   In this example, the two records might be in two different zones, and
   as a result might be administered by two different organizations, and
   signed by two different entities when using DNSSEC.  For these two
   reasons, using a prefix has recently become a very interesting
   solution for many protocol designers.  In some cases, e.g.,
   DomainKeys Identified Mail Signatures [<a href="./rfc4871" title="&quot;DomainKeys Identified Mail (DKIM) Signatures&quot;">RFC4871</a>], TXT records have
   been used.  In others, such as SRV, entirely new Resource Record
   Types have been added.

<span class="h3"><a class="selflink" id="section-3.3" href="#section-3.3">3.3</a>.  Add a Suffix to the Owner Name</span>

   Adding a suffix to a domain name changes the {owner, class, type}
   triple, and therefore the RRSet.  In this case, since the query name
   can be set to exactly the data one wants, the size of the RRSet is
   minimized.  The problem with adding a suffix is that it creates a
   parallel tree within the IN class.  Further, there is no technical
   mechanism to ensure that the delegation for "example.com" and
   "example.com._bar" are made to the same organization.  Furthermore,
   data associated with a single entity will now be stored in two
   different zones, such as "example.com" and "example.com._bar", which,
   depending on who controls "_bar", can create new synchronization and
   update authorization issues.

   One way of solving the administrative issues is by using the DNAME
   Resource Record Type specified in <a href="./rfc2672">RFC 2672</a> [<a href="./rfc2672" title="&quot;Non-Terminal DNS Name Redirection&quot;">RFC2672</a>].

   Even when using a different name, the data will still have to be
   stored in some Resource Record Type that has an appropriate format to
   store the data.  This implies that one might have to mix the prefix
   based selection mechanism with some other mechanism so that the right
   Resource Record can be found out of many in a potential larger RRSet.

   In <a href="./rfc2163">RFC 2163</a> [<a href="./rfc2163" title="&quot;Using the Internet DNS to Distribute MIXER Conformant Global Address Mapping (MCGAM)&quot;">RFC2163</a>] an infix token is inserted directly below the
   Top-Level Domain (TLD), but the result is equivalent to adding a
   suffix to the owner name (instead of creating a TLD, one is creating
   a second level domain).







<span class="grey">IAB, et al.                  Informational                      [Page 7]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-8" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


<span class="h3"><a class="selflink" id="section-3.4" href="#section-3.4">3.4</a>.  Add a New Class</span>

   DNS zones are class-specific in the sense that all the records in
   that zone share the same class as the zone's SOA record and the
   existence of a zone in one class does not guarantee the existence of
   the zone in any other class.  In practice, only the IN class has ever
   seen widespread deployment, and the administrative overhead of
   deploying an additional class would almost certainly be prohibitive.

   Nevertheless, one could, in theory, use the DNS class mechanism to
   distinguish between different kinds of data.  However, since the DNS
   delegation tree (represented by NS Resource Records) is itself tied
   to a specific class, attempting to resolve a query by crossing a
   class boundary may produce unexpected results because there is no
   guarantee that the name servers for the zone in the new class will be
   the same as the name servers in the IN class.  The MIT Hesiod system
   [<a href="#ref-Dyer87" title="&quot;Hesiod, Project Athena Technical Plan - Name Service&quot;">Dyer87</a>] used a scheme like this for storing data in the HS class,
   but only on a very small scale (within a single institution), and
   with an administrative fiat requiring that the delegation trees for
   the IN and HS trees be identical.  The use of the HS class for such
   storage of non-sensitive data was, over time, replaced by use of the
   Lightweight Directory Access Protocol (LDAP) [<a href="./rfc4511" title="&quot;Lightweight Directory Access Protocol (LDAP): The Protocol&quot;">RFC4511</a>].

   Even when using a different class, the data will still have to be
   stored in some Resource Record Type that has an appropriate format.

<span class="h3"><a class="selflink" id="section-3.5" href="#section-3.5">3.5</a>.  Add a New Resource Record Type</span>

   When adding a new Resource Record Type to the system, entities in
   four different roles have to be able to handle the new Type:

   1.  There must be a way to insert the new Resource Records into the
       zone at the Primary Master name server.  For some server
       implementations, the user interface only accepts Resource Record
       Types that it understands (perhaps so that the implementation can
       attempt to validate the data).  Other implementations allow the
       zone administrator to enter an integer for the Resource Record
       Type code and the RDATA in Base64 or hexadecimal encoding (or
       even as raw data).  <a href="./rfc3597">RFC 3597</a> [<a href="./rfc3597" title="&quot;Handling of Unknown DNS Resource Record (RR) Types&quot;">RFC3597</a>] specifies a standard
       generic encoding for this purpose.

   2.  A slave authoritative name server must be able to do a zone
       transfer, receive the data from some other authoritative name
       server, and serve data from the zone even though the zone
       includes records of unknown Resource Record Types.  Historically,
       some implementations have had problems parsing stored copies of
       the zone file after restarting, but those problems have not been
       seen for a few years.  Some implementations use an alternate



<span class="grey">IAB, et al.                  Informational                      [Page 8]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-9" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


       mechanism (e.g., LDAP) to transfer Resource Records in a zone,
       and are primarily used within corporate environments; in this
       case, name servers must be able to transfer new Resource Record
       Types using whatever mechanism is used.  However, today this
       alternative mechanism may not support unknown Resource Record
       Types.  Hence, in Internet environments, unknown Resource Record
       Types are supported, but in corporate environments they are
       problematic.

   3.  A caching resolver (most commonly a recursive name server) will
       cache the records that are responses to queries.  As mentioned in
       <a href="./rfc3597">RFC 3597</a> [<a href="./rfc3597" title="&quot;Handling of Unknown DNS Resource Record (RR) Types&quot;">RFC3597</a>], there are various pitfalls where a recursive
       name server might end up having problems.

   4.  The application must be able to get the RRSet with a new Resource
       Record Type.  The application itself may understand the RDATA,
       but the resolver library might not.  Support for a generic
       interface for retrieving arbitrary DNS Resource Record Types has
       been a requirement since 1989 (see <a href="./rfc1123#section-6.1.4.2">Section&nbsp;6.1.4.2 of [RFC1123]</a>).
       Some stub resolver library implementations neglect to provide
       this functionality and cannot handle unknown Resource Record
       Types, but implementation of a new stub resolver library is not
       particularly difficult, and open source libraries that already
       provide this functionality are available.

   Historically, adding a new Resource Record Type has been very
   problematic.  The review process has been cumbersome, DNS servers
   have not been able to handle new Resource Record Types, and firewalls
   have dropped queries or responses with Resource Record Types that are
   unknown to the firewall.  This is, for example, one of the reasons
   the ENUM standard reuses the NAPTR Resource Record, a decision that
   today might have gone to creating a new Resource Record Type instead.

   Today, there is a requirement that DNS software handle unknown
   Resource Record Types, and investigations have shown that software
   that is deployed, in general, does support it, except in some
   alternate mechanisms for transferring Resource Records such as LDAP,
   as noted above.  Also, the approval process for new Resource Record
   Types has been updated [<a href="./rfc5395" title="&quot;Domain Name System (DNS) IANA Considerations&quot;">RFC5395</a>] so the effort that is needed for
   various Resource Record Types is more predictable.

<span class="h2"><a class="selflink" id="section-4" href="#section-4">4</a>.  Zone Boundaries are Invisible to Applications</span>

   Regardless of the possible choices above, we have seen a number of
   cases where the application made assumptions about the structure of
   the namespace and the location where specific information resides.
   We take a small sidestep to argue against such approaches.




<span class="grey">IAB, et al.                  Informational                      [Page 9]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-10" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


   The DNS namespace is a hierarchy, technically speaking.  However,
   this only refers to the way names are built from multiple labels.
   DNS hierarchy neither follows nor implies administrative hierarchy.
   Because of that, it cannot be assumed that data attached to a node in
   the DNS tree is valid for the whole subtree.  Technically, there are
   zone boundaries partitioning the namespace, and administrative
   boundaries (or policy boundaries) may even exist elsewhere.

   The false assumption has lead to an approach called "tree climbing",
   where a query that does not receive a positive response (either the
   requested RRSet was missing or the name did not exist) is retried by
   repeatedly stripping off the leftmost label (climbing towards the
   root) until the root domain is reached.  Sometimes these proposals
   try to avoid the query for the root or the TLD level, but still this
   approach has severe drawbacks:

   o  Technically, the DNS was built as a query-response tool without
      any search capability [<a href="./rfc3467" title="&quot;Role of the Domain Name System (DNS)&quot;">RFC3467</a>].  Adding the search mechanism
      imposes additional burden on the technical infrastructure, in the
      worst case on TLD and root name servers.

   o  For reasons similar to those outlined in <a href="./rfc1535">RFC 1535</a> [<a href="./rfc1535" title="&quot;A Security Problem and Proposed Correction With Widely Deployed DNS Software&quot;">RFC1535</a>],
      querying for information in a domain outside the control of the
      intended entity may lead to incorrect results and may also put
      security at risk.  Finding the exact policy boundary is impossible
      without an explicit marker, which does not exist at present.  At
      best, software can detect zone boundaries (e.g., by looking for
      SOA Resource Records), but some TLD registries register names
      starting at the second level (e.g., CO.UK), and there are various
      other "registry" types at second, third, or other level domains
      that cannot be identified as such without policy knowledge
      external to the DNS.

   To restate, the zone boundary is purely a boundary that exists in the
   DNS for administrative purposes, and applications should be careful
   not to draw unwarranted conclusions from zone boundaries.  A
   different way of stating this is that the DNS does not support
   inheritance, e.g., an MX RRSet for a TLD will not be valid for any
   subdomain of that particular TLD.

<span class="h2"><a class="selflink" id="section-5" href="#section-5">5</a>.  Why Adding a New Resource Record Type Is the Preferred Solution</span>

   By now, the astute reader might be wondering what conclusions to draw
   from the issues presented so far.  We will now attempt to clear up
   the reader's confusion by following the thought processes of a
   typical application designer who wishes to store data in the DNS.
   We'll show how such a designer almost inevitably hits upon the idea
   of just using a TXT Resource Record, why this is a bad thing, and why



<span class="grey">IAB, et al.                  Informational                     [Page 10]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-11" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


   a new Resource Record Type should be allocated instead.  We'll also
   explain how the reuse of an existing Resource Record, including TXT,
   can be made less harmful.

   The overall problem with most solutions has to do with two main
   issues:

   o  No semantics to prevent collision with other use

   o  Space considerations in the DNS message

   A typical application designer is not interested in the DNS for its
   own sake, but rather regards it as a distributed database in which
   application data can be stored.  As a result, the designer of a new
   application is usually looking for the easiest way to add whatever
   new data the application needs to the DNS in a way that naturally
   associates the data with a DNS name and does not require major
   changes to DNS servers.

   As explained in <a href="#section-3.4">Section 3.4</a>, using the DNS class system as an
   extension mechanism is not really an option, and in fact, most users
   of the system don't even realize that the mechanism exists.  As a
   practical matter, therefore any extension is likely to be within the
   IN class.

   Adding a new Resource Record Type is the technically correct answer
   from the DNS protocol standpoint (more on this below), but doing so
   requires some DNS expertise, due to the issues listed in <a href="#section-3.5">Section 3.5</a>.
   Consequently, this option is often rejected.  Note that according to
   <a href="./rfc5395">RFC 5395</a> [<a href="./rfc5395" title="&quot;Domain Name System (DNS) IANA Considerations&quot;">RFC5395</a>], some Types require IETF Consensus, while others
   only require a specification.

   There is a drawback to defining new RR types that is worth
   mentioning.  The Resource Record Type (RRTYPE) is a 16-bit value and
   hence is a limited resource.  In order to prevent hoarding the
   registry has a review-based allocation policy [<a href="./rfc5395" title="&quot;Domain Name System (DNS) IANA Considerations&quot;">RFC5395</a>]; however,
   this may not be sufficient if extension of the DNS by addition of new
   RR types takes up significantly and the registry starts nearing
   completion.  In that case, the trade-offs with respect to choosing an
   extension mechanism may need to change.

   The application designer is thus left with the prospect of reusing
   some existing DNS Types within the IN class, but when the designer
   looks at the existing Types, almost all of them have well-defined
   semantics, none of which quite match the needs of the new
   application.  This has not completely prevented proposals from





<span class="grey">IAB, et al.                  Informational                     [Page 11]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-12" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


   reusing existing Resource Record Types in ways incompatible with
   their defined semantics, but it does tend to steer application
   designers away from this approach.

   For example, Resource Record Type 40 was registered for the SINK
   Resource Record Type.  This Resource Record Type was discussed in the
   DNSIND working group of the IETF, and it was decided at the 46th IETF
   to not move the I-D forward to become an RFC because of the risk of
   encouraging application designers to use the SINK Resource Record
   Type instead of registering a new Resource Record Type, which would
   result in infeasibly large SINK RRsets.

   Eliminating all of the above leaves the TXT Resource Record Type in
   the IN class.  The TXT RDATA format is free form text, and there are
   no existing semantics to get in the way.  Some attempts have been
   made, for example, in [<a href="#ref-DNSEXT-DNS-SD" title="&quot;DNS-Based Service Discovery&quot;">DNSEXT-DNS-SD</a>], to specify a structured format
   for TXT Resource Record Types, but no such attempt has reached RFC
   status.  Furthermore, the TXT Resource Record can obviously just be
   used as a bucket in which to carry around data to be used by some
   higher-level parser, perhaps in some human-readable programming or
   markup language.  Thus, for many applications, TXT Resource Records
   are the "obvious" choice.  Unfortunately, this conclusion, while
   understandable, is also problematic, for several reasons.

   The first reason why TXT Resource Records are not well suited to such
   use is precisely what makes them so attractive: the lack of pre-
   defined common syntax or structure.  As a result, each application
   that uses them creates its own syntax/structure, and that makes it
   difficult to reliably distinguish one application's record from
   others, and for its parser to avoid problems when it encounters other
   TXT records.

   Arguably, the TXT Resource Record is misnamed, and should have been
   called the Local Container record, because a TXT Resource Record
   means only what the data producer says it means.  This is fine, so
   long as TXT Resource Records are being used by human beings or by
   private agreement between data producer and data consumer.  However,
   it becomes a problem once one starts using them for standardized
   protocols in which there is no prior relationship between data
   producer and data consumer.  If TXT records are used without one of
   the naming modifications discussed earlier (and in some cases even if
   one uses such naming mechanisms), there is nothing to prevent
   collisions with some other incompatible use of TXT Resource Records.

   This is even worse than the general subtyping problem described in
   <a href="#section-3.1">Section 3.1</a> because TXT Resource Records don't even have a
   standardized selector field in which to store the subtype.  <a href="./rfc1464">RFC 1464</a>
   [<a href="./rfc1464" title="&quot;Using the Domain Name System To Store Arbitrary String Attributes&quot;">RFC1464</a>] tried, but it was not a success.  At best, a definition of



<span class="grey">IAB, et al.                  Informational                     [Page 12]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-13" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


   a subtype is reduced to hoping that whatever scheme one has come up
   with will not accidently conflict with somebody else's subtyping
   scheme, and that it will not be possible to mis-parse one
   application's use of TXT Resource Records as data intended for a
   different application.  Any attempt to impose a standardized format
   within the TXT Resource Record format would be at least fifteen years
   too late, even if it were put into effect immediately; at best, one
   can restrict the syntax that a particular application uses within a
   TXT Resource Record and accept the risk that unrelated TXT Resource
   Record uses will collide with it.

   Using one of the naming modifications discussed in <a href="#section-3.2">Section 3.2</a> and
   <a href="#section-3.3">Section 3.3</a> would address the subtyping problem, (and have been used
   in combinations with reuse of TXT record, such as for the dns/txt
   lookup mechanism in Domain Keys Identified Mail (DKIM)) but each of
   these approaches brings in new problems of its own.  The prefix
   approach (that for example SRV Resource Records use) does not work
   well with wildcards, which is a particular problem for mail-related
   applications, since MX Resource Records are probably the most common
   use of DNS wildcards.  The suffix approach doesn't have wildcard
   issues, but, as noted previously, it does have synchronization and
   update authorization issues, since it works by creating a second
   subtree in a different part of the global DNS namespace.

   The next reason why TXT Resource Records are not well suited to
   protocol use has to do with the limited data space available in a DNS
   message.  As alluded to briefly in <a href="#section-3.1">Section 3.1</a>, typical DNS query
   traffic patterns involve a very large number of DNS clients sending
   queries to a relatively small number of DNS servers.  Normal path MTU
   discovery schemes do little good here because, from the server's
   perspective, there isn't enough repeat traffic from any one client
   for it to be worth retaining state.  UDP-based DNS is an idempotent
   query, whereas TCP-based DNS requires the server to keep state (in
   the form of TCP connection state, usually in the server's kernel) and
   roughly triples the traffic load.  Thus, there's a strong incentive
   to keep DNS messages short enough to fit in a UDP datagram,
   preferably a UDP datagram short enough not to require IP
   fragmentation.

   Subtyping schemes are therefore again problematic because they
   produce larger Resource RRSets than necessary, but verbose text
   encodings of data are also wasteful since the data they hold can
   usually be represented more compactly in a Resource Record designed
   specifically to support the application's particular data needs.  If
   the data that need to be carried are so large that there is no way to
   make them fit comfortably into the DNS regardless of encoding, it is
   probably better to move the data somewhere else, and just use the DNS
   as a pointer to the data, as with NAPTR.



<span class="grey">IAB, et al.                  Informational                     [Page 13]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-14" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


<span class="h2"><a class="selflink" id="section-6" href="#section-6">6</a>.  Conclusion and Recommendation</span>

   Given the problems detailed in <a href="#section-5">Section 5</a>, it is worth reexamining the
   oft-jumped-to conclusion that specifying a new Resource Record Type
   is hard.  Historically, this was indeed the case, but recent surveys
   suggest that support for unknown Resource Record Types [<a href="./rfc3597" title="&quot;Handling of Unknown DNS Resource Record (RR) Types&quot;">RFC3597</a>] is
   now widespread in the public Internet, and because of that, the DNS
   infrastructure can handle new Resource Record Types.  The lack of
   support for unknown Types remains an issue for relatively old
   provisioning software and in corporate environments.

   Of all the issues detailed in <a href="#section-3.5">Section 3.5</a>, provisioning the data is
   in some respects the most difficult.  Investigations with zone
   transfers show that the problem is less difficult for the
   authoritative name servers themselves than the front-end systems used
   to enter (and perhaps validate) the data.  Hand editing does not work
   well for maintenance of large zones, so some sort of tool is
   necessary, and the tool may not be tightly coupled to the name server
   implementation itself.  Note, however, that this provisioning problem
   exists to some degree with any new form of data to be stored in the
   DNS, regardless of data format, Resource Record type (even if TXT
   Resource Record Types are in use), or naming scheme.  Adapting front-
   end systems to support a new Resource Record Type may be a bit more
   difficult than reusing an existing type, but this appears to be a
   minor difference in degree rather than a difference in kind.

   Given the various issues described in this note, we believe that:

   o  there is no magic solution that allows a completely painless
      addition of new data to the DNS, but

   o  on the whole, the best solution is still to use the DNS Resource
      Record Type mechanism designed for precisely this purpose,
      whenever possible, and

   o  of all the alternate solutions, the "obvious" approach of using
      TXT Resource Records for arbitrary names is almost certainly the
      worst, especially for the two reasons outlined above (lack of
      semantics and its implementations, and size leading to the need to
      use TCP).

<span class="h2"><a class="selflink" id="section-7" href="#section-7">7</a>.  Creating a New Resource Record Type</span>

   The process for creating a new Resource Record Type is specified in
   <a href="./rfc5395">RFC 5395</a> [<a href="./rfc5395" title="&quot;Domain Name System (DNS) IANA Considerations&quot;">RFC5395</a>].






<span class="grey">IAB, et al.                  Informational                     [Page 14]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-15" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


<span class="h2"><a class="selflink" id="section-8" href="#section-8">8</a>.  Security Considerations</span>

   DNS RRSets can be signed using DNSSEC.  DNSSEC is almost certainly
   necessary for any application mechanism that stores authorization
   data in the DNS.  DNSSEC signatures significantly increase the size
   of the messages transported, and because of this, the DNS message
   size issues discussed in Sections <a href="#section-3.1">3.1</a> and <a href="#section-5">5</a> are more serious than
   they might at first appear.

   Adding new Resource Record Types (as discussed in <a href="#section-3.5">Section 3.5</a>) can
   create two different kinds of problems: in the DNS software and in
   applications.  In the DNS software, it might conceivably trigger bugs
   and other bad behavior in software that is not compliant with <a href="./rfc3597">RFC</a>
   <a href="./rfc3597">3597</a> [<a href="./rfc3597" title="&quot;Handling of Unknown DNS Resource Record (RR) Types&quot;">RFC3597</a>], but most such DNS software is old enough and insecure
   enough that it should be updated for other reasons in any case.  In
   applications and provisioning software, the changes for the new
   features that need the new data in the DNS can be updated to
   understand the structure of the new data format (regardless of
   whether a new Resource Record Type is used or some other mechanism is
   chosen).  Basic API support for retrieving arbitrary Resource Record
   Types has been a requirement since 1989 [<a href="./rfc1123" title="&quot;Requirements for Internet Hosts - Application and Support&quot;">RFC1123</a>].

   Any new protocol that proposes to use the DNS to store data used to
   make authorization decisions would be well advised not only to use
   DNSSEC but also to encourage upgrades to DNS server software recent
   enough not to be riddled with well-known exploitable bugs.

<span class="h2"><a class="selflink" id="section-9" href="#section-9">9</a>.  Acknowledgements</span>

   This document has been created over a number of years, with input
   from many people.  The question on how to expand and use the DNS is
   sensitive, and a document like this can not please everyone.  The
   goal is instead to describe the architecture and tradeoffs, and make
   some recommendations about best practices.

   People that have helped include: Dean Anderson, Mark Andrews, John
   Angelmo, Roy Badami, Dan Bernstein, Alex Bligh, Nathaniel Borenstein,
   Stephane Bortzmeyer, Brian Carpenter, Leslie Daigle, Elwyn Davies,
   Mark Delany, Richard Draves, Martin Duerst, Donald Eastlake, Robert
   Elz, Jim Fenton, Tony Finch, Jim Gilroy, Olafur Gudmundsson, Eric
   Hall, Phillip Hallam-Baker, Ted Hardie, Bob Hinden, Paul Hoffman,
   Geoff Houston, Christian Huitema, Johan Ihren, John Klensin, Ben
   Laurie, William Leibzon, John Levine, Edward Lewis, David MacQuigg,
   Allison Mankin, Bill Manning, David Meyer, Pekka Nikander, Mans
   Nilsson, Masataka Ohta, Douglas Otis, Michael Patton, Jonathan
   Rosenberg, Anders Rundgren, Miriam Sapiro, Carsten Strotmann, Pekka
   Savola, Chip Sharp, James Snell, Michael Thomas, Paul Vixie, Sam
   Weiler, Florian Weimer, Bert Wijnen, and Dan Wing.



<span class="grey">IAB, et al.                  Informational                     [Page 15]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-16" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


<span class="h2"><a class="selflink" id="section-10" href="#section-10">10</a>.  IAB Members at the Time of This Writing</span>

   Loa Andersson
   Gonzalo Camarillo
   Stuart Cheshire
   Russ Housley
   Olaf Kolkman
   Gregory Lebovitz
   Barry Leiba
   Kurtis Lindqvist
   Andrew Malis
   Danny McPherson
   David Oran
   Dave Thaler
   Lixia Zhang

<span class="h2"><a class="selflink" id="section-11" href="#section-11">11</a>.  References</span>

<span class="h3"><a class="selflink" id="section-11.1" href="#section-11.1">11.1</a>.  Normative References</span>

   [<a id="ref-RFC1035">RFC1035</a>]        Mockapetris, P., "Domain names - implementation and
                    specification", STD 13, <a href="./rfc1035">RFC 1035</a>, November 1987.

   [<a id="ref-RFC1464">RFC1464</a>]        Rosenbaum, R., "Using the Domain Name System To
                    Store Arbitrary String Attributes", <a href="./rfc1464">RFC 1464</a>,
                    May 1993.

   [<a id="ref-RFC2535">RFC2535</a>]        Eastlake, D., "Domain Name System Security
                    Extensions", <a href="./rfc2535">RFC 2535</a>, March 1999.

   [<a id="ref-RFC2671">RFC2671</a>]        Vixie, P., "Extension Mechanisms for DNS (EDNS0)",
                    <a href="./rfc2671">RFC 2671</a>, August 1999.

   [<a id="ref-RFC3597">RFC3597</a>]        Gustafsson, A., "Handling of Unknown DNS Resource
                    Record (RR) Types", <a href="./rfc3597">RFC 3597</a>, September 2003.

   [<a id="ref-RFC5395">RFC5395</a>]        Eastlake, D., "Domain Name System (DNS) IANA
                    Considerations", <a href="https://www.rfc-editor.org/bcp/bcp42">BCP 42</a>, <a href="./rfc5395">RFC 5395</a>, November 2008.

<span class="h3"><a class="selflink" id="section-11.2" href="#section-11.2">11.2</a>.  Informative References</span>

   [<a id="ref-DNSEXT-DNS-SD">DNSEXT-DNS-SD</a>]  Cheshire, S. and M. Krochmal, "DNS-Based Service
                    Discovery", Work in Progress, September 2008.

   [<a id="ref-Dyer87">Dyer87</a>]         Dyer, S. and F. Hsu, "Hesiod, Project Athena
                    Technical Plan - Name Service", Version 1.9,
                    April 1987.




<span class="grey">IAB, et al.                  Informational                     [Page 16]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-17" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


   [<a id="ref-RFC1123">RFC1123</a>]        Braden, R., "Requirements for Internet Hosts -
                    Application and Support", STD 3, <a href="./rfc1123">RFC 1123</a>,
                    October 1989.

   [<a id="ref-RFC1535">RFC1535</a>]        Gavron, E., "A Security Problem and Proposed
                    Correction With Widely Deployed DNS Software",
                    <a href="./rfc1535">RFC 1535</a>, October 1993.

   [<a id="ref-RFC2163">RFC2163</a>]        Allocchio, C., "Using the Internet DNS to Distribute
                    MIXER Conformant Global Address Mapping (MCGAM)",
                    <a href="./rfc2163">RFC 2163</a>, January 1998.

   [<a id="ref-RFC2181">RFC2181</a>]        Elz, R. and R. Bush, "Clarifications to the DNS
                    Specification", <a href="./rfc2181">RFC 2181</a>, July 1997.

   [<a id="ref-RFC2672">RFC2672</a>]        Crawford, M., "Non-Terminal DNS Name Redirection",
                    <a href="./rfc2672">RFC 2672</a>, August 1999.

   [<a id="ref-RFC3445">RFC3445</a>]        Massey, D. and S. Rose, "Limiting the Scope of the
                    KEY Resource Record (RR)", <a href="./rfc3445">RFC 3445</a>, December 2002.

   [<a id="ref-RFC3467">RFC3467</a>]        Klensin, J., "Role of the Domain Name System (DNS)",
                    <a href="./rfc3467">RFC 3467</a>, February 2003.

   [<a id="ref-RFC3761">RFC3761</a>]        Faltstrom, P. and M. Mealling, "The E.164 to Uniform
                    Resource Identifiers (URI) Dynamic Delegation
                    Discovery System (DDDS) Application (ENUM)",
                    <a href="./rfc3761">RFC 3761</a>, April 2004.

   [<a id="ref-RFC4033">RFC4033</a>]        Arends, R., Austein, R., Larson, M., Massey, D., and
                    S. Rose, "DNS Security Introduction and
                    Requirements", <a href="./rfc4033">RFC 4033</a>, March 2005.

   [<a id="ref-RFC4034">RFC4034</a>]        Arends, R., Austein, R., Larson, M., Massey, D., and
                    S. Rose, "Resource Records for the DNS Security
                    Extensions", <a href="./rfc4034">RFC 4034</a>, March 2005.

   [<a id="ref-RFC4035">RFC4035</a>]        Arends, R., Austein, R., Larson, M., Massey, D., and
                    S. Rose, "Protocol Modifications for the DNS
                    Security Extensions", <a href="./rfc4035">RFC 4035</a>, March 2005.

   [<a id="ref-RFC4511">RFC4511</a>]        Sermersheim, J., "Lightweight Directory Access
                    Protocol (LDAP): The Protocol", <a href="./rfc4511">RFC 4511</a>, June 2006.

   [<a id="ref-RFC4592">RFC4592</a>]        Lewis, E., "The Role of Wildcards in the Domain Name
                    System", <a href="./rfc4592">RFC 4592</a>, July 2006.





<span class="grey">IAB, et al.                  Informational                     [Page 17]</span></pre>
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-18" ></span>
<span class="grey"><a href="./rfc5507">RFC 5507</a>         Design Choices When Expanding the DNS        April 2009</span>


   [<a id="ref-RFC4871">RFC4871</a>]        Allman, E., Callas, J., Delany, M., Libbey, M.,
                    Fenton, J., and M. Thomas, "DomainKeys Identified
                    Mail (DKIM) Signatures", <a href="./rfc4871">RFC 4871</a>, May 2007.

Authors' Addresses

   Internet Architecture Board

   EMail: iab@iab.org


   Patrik Faltstrom (editor)

   EMail: paf@cisco.com


   Rob Austein (editor)

   EMail: sra@isc.org


   Peter Koch (editor)

   EMail: pk@denic.de



























IAB, et al.                  Informational                     [Page 18]
</pre>