Internet-Draft | dns+cbor | November 2023 |
Lenders, et al. | Expires 20 May 2024 | [Page] |
This document specifies a compressed data format of DNS messages using the Concise Binary Object Representation [RFC8949]. The primary purpose is to keep DNS messages small in constrained networks.¶
This note is to be removed before publishing as an RFC.¶
The latest revision of this draft can be found at https://anr-bmbf-pivot.github.io/draft-lenders-dns-cbor/draft-lenders-dns-cbor.html. Status information for this document may be found at https://datatracker.ietf.org/doc/draft-lenders-dns-cbor/.¶
Discussion of this document takes place on the CBOR Working Group mailing list (mailto:cbor@ietf.org), which is archived at https://mailarchive.ietf.org/arch/browse/cbor/. Subscribe at https://www.ietf.org/mailman/listinfo/cbor/.¶
Source for this draft and an issue tracker can be found at https://github.com/anr-bmbf-pivot/draft-lenders-dns-cbor.¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 20 May 2024.¶
Copyright (c) 2023 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
In constrained networks [RFC7228], the link layer may restrict the payload sizes to only a few hundreds bytes. Encrypted DNS resolution, such as DNS over HTTPS (DoH) [RFC8484] or DNS over CoAP (DoC) [I-D.ietf-core-dns-over-coap], may lead to DNS message sizes that exceed this limit, even when implementing header compression such as 6LoWPAN IPHC [RFC6282] or SCHC [RFC8724], [RFC8824].¶
Although adoption layers such as 6LoWPAN [RFC4944] or SCHC [RFC8724] offer fragmentation to comply with small MTUs, fragmentation should be avoided in constrained networks, because fragmentation combined with high packet loss multiplies the loss. As such, a compression format for DNS messages is needed.¶
This document specifies a compressed data format for DNS messages. DNS messages are encoded in Concise Binary Object Representation (CBOR) [RFC8949] and, additionally, unnecessary or redundant information is removed. To use the outcome of this specification in DoH and DoC, this document also specifies a Media Type header for DoH and a Content-Format option for DoC.¶
CBOR types (unsigned integer, byte string, text string, arrays, etc.) are used as defined in [RFC8949].¶
TBD DNS server and client.¶
A DNS query is a message that queries DNS information from an upstream DNS resolver.¶
The term "constrained networks" is used as defined in [RFC7228].¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
To keep overhead minimal, a DNS message is represented as CBOR arrays. All CBOR items used in this specification are of definite length. CBOR arrays that do not follow the length definitions of this or follow-up specifications, MUST be silently ignored. It is assumed that DNS query and DNS response are distinguished message types and that the query can be mapped to the response by the transfer protocol of choice. To define the representation of binary objects we use the Concise Data Definition Language (CDDL) [RFC8610].¶
If, for any reason, a DNS message is not representable in the CBOR format specified in this document, a fallback to the another DNS message format, e.g., the classic DNS wire format, MUST always be possible.¶
Domain names are represented in their commonly known string format (e.g., "example.org", see Section 2.3.1 in [RFC1035]) and in IDNA encoding [RFC5890] as a text string. For the purpose of this document, domain names remain case-insensitive as specified in [RFC1035].¶
The representation of a domain name is defined in Figure 2.¶
TBD: represent names as components ((* tstr)
), provide name compression when [I-D.ietf-cbor-packed] is
updated for the reference format and table building discussed at IETF 118.¶
This document specifies the representation of both standard DNS resource records (RRs, see [RFC1035]) and EDNS option pseudo-RRs (see [RFC6891]). If for any reason, a resource record can not be represented in the given formats, they can be represented in their binary wire-format form, as a byte string.¶
Further special records, e.g., TSIG can be defined in follow-up specifications and are out of scope of this document.¶
The representation of a DNS resource records is defined in Figure 3.¶
Standard DNS resource records are encoded as CBOR arrays containing 2 to 5 entries in the following order:¶
An optional name (as text string, see Section 3.1),¶
A TTL (as unsigned integer),¶
An optional record type (as unsigned integer),¶
An optional record class (as unsigned integer), and lastly¶
A record data entry (as unsigned integer, negative integer, byte string, or text string).¶
If the first item of the resource record is a text string, it is its name. If the name is elided, the name is derived from the question section of the message. For responses, the question section is either taken from the query (see Section 3.3) or provided with the response see Section 3.4. The query may be derived from the context of the transfer protocol.¶
If the record type is elided, the record type from the question is assumed. If record class is elided, the record class from the question is assumed. When a record class is required, the record type MUST also be provided.¶
The byte format of the record data as a byte string follows the wire format as specified in Section 3.3 [RFC1035] (or other specifications of the respective record type). Note that this format does not include the RDLENGTH field from [RFC1035] as this value is encoded in the length field of the CBOR byte string.¶
If the record data represents a domain name (e.g., for CNAME or PTR records), the record data MAY be represented as a text string as specified in Section 3.1. This can save 1 byte of data, because the byte representation of DNS names requires both an additional byte to define the length of the first name component and well as a zero byte at the end of the name. With CBOR on the other hand only 1 byte is required to define type and length of the text string up until a string length of 23 characters.¶
There is an argument to be made for more structured formats of other record data representations (e.g. MX or SOA), but these usually add more overhead. As such, those record data are to be represented as a byte string.¶
EDNS OPT Pseudo-RRs are represented as a CBOR array. To distinguish them from normal standard RRs, they are marked with tag TBD141.¶
Name and record type can be elided as they are always "." and OPT (41), respectively [RFC6891].¶
The UDP payload size may be the first element as an unsigned integer in the array but it can be elided if it defaults to 512, the maximum allowable size for DNS over UDP [RFC6891].¶
The next element is an array of the options, which are represented two elements each, an unsigned integer, the option code, followed by a byte string, the option data. Multiple options alternate between unsigned integer and byte string within the array.¶
After that, up to three unsigned integers are following. The first being the extended flags as unsigned integer (implied to be 0 if elided), the second the extended RCODE as an unsigned integer (implied to be 0 if elided), and the third the EDNS version (implied to be 0 if elided). They are dependent on each of their previous elements. If the EDNS version is not elided, both extended flags and extended RCODE MUST not be elided. If the RCODE is not elided the extended flags MUST not be elided.¶
TBD: reverse extended flags to get MSB-defined DO into LSB?¶
Note that future EDNS versions may require a different format than the one described above.¶
DNS queries are encoded as CBOR arrays containing up to 5 entries in the following order:¶
An optional flag field (as unsigned integer),¶
The question section (as array),¶
An optional authority section (as array), and¶
An optional additional section (as array)¶
If the first item of the query is an array, it is the question section, if it is an unsigned integer, it is as flag field and maps to the header flags in [RFC1035] and the "DNS Header Flags" IANA registry including the QR flag and the Opcode. It MUST be lesser than 2^16.¶
If the flags are elided, the value 0 is assumed.¶
This specification assumes that the DNS messages are sent over a transfer protocol that can map the queries to their responses, e.g., DNS over HTTPS [RFC8484] or DNS over CoAP [I-D.ietf-core-dns-over-coap]. As a consequence, the DNS transaction ID is always elided and the value 0 is assumed.¶
A question within the question section is encoded as a CBOR array containing up to 3 entries:¶
The queried name (as text string, see Section 3.1),¶
An optional record type (as unsigned integer), and¶
An optional record class (as unsigned integer)¶
If the record type is elided, record type AAAA
as specified in [RFC3596] is assumed.
If the record class is elided, record class IN
as specified in [RFC1035] is assumed.
When a record class is required, the record type MUST also be provided.¶
If more than one question is supposed to be in the question section, the next question just follows. In this case, for every question but the record type MUST be included and it is not optional. This way it is ensured that the parser can distinguish each question by looking up the name first (TBD note: this is especially relevant once the name is split up in components).¶
The remainder of the query is either empty or MUST consist of up to two arrays. The first array, if present, encodes the authority section of the query as an array of DNS resource records (see Section 3.2) The second array, if present, encodes the additional section of the query as an array of DNS resource records (see Section 3.2)¶
The representation of a DNS query is defined in Figure 6.¶
DNS responses are encoded as a CBOR array containing up to 7 entries.¶
An optional flag field (as unsigned integer),¶
An optional question section (as array, encoded as described in Section 3.3)¶
The answer section (as array),¶
An optional authority section (as array), and¶
An optional additional section (as array)¶
As for queries, the DNS transaction ID is elided and implied to be 0.¶
If the CBOR array is a response to a query for which the flags indicate that flags are set in the response, they MUST be set accordingly and thus included in the response. If the flags are not included, the flags are implied to be 0x8000 (everything unset except for the QR flag).¶
If the response includes only 1 array, this is the DNS answer section represented as an array of one or more DNS Resource Records (see Section 3.2).¶
If the response includes more than 2 arrays, the first entry may be the question section, identified by not being an array of arrays. If it is present, it is followed by the answer section. The question section is encoded as specified in Section 3.3.¶
If the answer section is followed by 1 additional array, it is the additional section (TBD: back choice to favor additional section by empirical data). Like the answer section, the additional sections is represented as an array of one or more DNS Resource Records (see Section 3.2).¶
If the answer section is followed by 2 additional arrays, the first is the authority section, and the second the additional section (TBD: back choice to favor additional section by empirical data). The authority section is also represented as an array of one or more DNS Resource Records (see Section 3.2).¶
If both DNS server and client support CBOR-packed [I-D.ietf-cbor-packed], it MAY be used for name and address compression in DNS responses.¶
A DNS client uses media type "application/dns+cbor;packed=1" to negotiate (see, e.g., [RFC9110] or [RFC7252], Section 5.5.4) with the DNS server if the server supports packed CBOR. If it does, it MAY request the response to be in CBOR-packed (media type "applicaton/dns+cbor;packed=1"). The server then SHOULD reply with the response in CBOR-packed.¶
The representation of DNS responses in CBOR-packed has the same semantics as for tag TBD113 ([I-D.ietf-cbor-packed], Section 3.1) with the rump being the compressed response. The difference to [I-D.ietf-cbor-packed] is that tag TBD113 is OPTIONAL.¶
Packed compression of queries is not specified, as apart from EDNS(0) (see Section 3.2.2), they only consist of one question most of the time.¶
How the compressor constructs the packing table, i.e., how the compression is applied, is out of scope of this document. Several potential compression algorithms were evaluated in [TBD].¶
TBD: Table comparing DNS wire-format, DNS+CBOR, and DNS+CBOR-packed¶
This section records the status of known implementations of the protocol defined by this specification at the time of posting of this Internet-Draft, and is based on a proposal described in [RFC7942]. The description of implementations in this section is intended to assist the IETF in its decision processes in progressing drafts to RFCs. Please note that the listing of any individual implementation here does not imply endorsement by the IETF. Furthermore, no effort has been spent to verify the information presented here that was supplied by IETF contributors. This is not intended as, and must not be construed to be, a catalog of available implementations or their features. Readers are advised to note that other implementations may exist.¶
According to [RFC7942], "this will allow reviewers and working groups to assign due consideration to documents that have the benefit of running code, which may serve as evidence of valuable experimentation and feedback that have made the implemented protocols more mature. It is up to the individual working groups to use this information as they see fit".¶
The authors of this document provide a decoder/encoder implementation of both the unpacked and packed format specified in this document in Python.¶
The authors of this document provide a decoder/encoder implementation of the unpacked format specified in this document for the RIOT operating system. It can only encode queries and decode responses.¶
TODO Security¶
This document registers a media type for the serialization format of DNS messages in CBOR. It follows the procedures specified in [RFC6838].¶
Type name: application¶
Subtype name: dns+cbor¶
Required parameters: None¶
Optional parameters: packed¶
Encoding considerations: Must be encoded as using [RFC8949]. See [TBD-this-spec] for details.¶
Security considerations: See Section 7 of this draft¶
Interoperability considerations: TBD¶
Published specification: [TBD-this-spec]¶
Applications that use this media type: TBD DNS over X systems¶
Fragment Identifier Considerations: TBD¶
Additional information:¶
Deprecated alias names for this type: N/A¶
Magic number(s): N/A¶
File extension(s): dnsc¶
Macintosh file type code(s): none¶
Person & email address to contact for further information: Martine S. Lenders m.lenders@fu-berlin.de¶
Intended usage: COMMON¶
Restrictions on Usage: None?¶
Author: Martine S. Lenders m.lenders@fu-berlin.de¶
Change controller: Martine S. Lenders m.lenders@fu-berlin.de¶
Provisional registrations? No¶
IANA is requested to assign CoAP Content-Format ID for the new DNS message media types in the "CoAP Content-Formats" sub-registry, within the "CoRE Parameters" registry [RFC7252], corresponding the "application/dns+cbor" media type specified in Section 8.1:¶
A DNS query of the record AAAA
in class IN
for name "example.org" is
represented in CBOR extended diagnostic notation (EDN) (see Section 8 in
[RFC8949] and Appendix G in [RFC8610]) as follows:¶
[["example.org"]]¶
A query of an A
record for the same name is represented as¶
[["example.org", 1]]¶
A query of ANY
record for that name is represented as¶
[["example.org", 255, 255]]¶
The responses to the examples provided in Appendix A.1 are shown below. We use the CBOR extended diagnostic notation (EDN) (see Section 8 in [RFC8949] and Appendix G in [RFC8610]).¶
To represent an AAAA
record with TTL 300 seconds for the IPv6 address 2001:db8::1, a minimal
response to ["example.org"]
could be¶
[[[300, h'20010db8000000000000000000000001']]]¶
In this case, the name is derived from the query.¶
If the name or the context is required, the following response would also be valid:¶
[[["example.org", 300, h'20010db8000000000000000000000001']]]¶
If the query can not be mapped to the response for some reason, a response would look like:¶
[["example.org"], [[300, h'20010db8000000000000000000000001']]]¶
To represent a minimal response of an A
record with TTL 3600 seconds for the IPv4 address
192.0.2.1, a minimal response to ["example.org", 1]
could be¶
[[300, h'c0000201']]¶
Note that here also the 1 of record type A
can be elided, as this record
type is specified in the question section.¶
Lastly, a response to ["example.org", 255, 255]
could be¶
[ ["example.org", 12, 1], [[3600, "_coap._udp.local"]], [ [3600, 2, "ns1.example.org"], [3600, 2, "ns2.example.org"] ], [ [ "_coap._udp.local", 3600, 28, h'20010db8000000000000000000000001' ], [ "_coap._udp.local", 3600, 28, h'20010db8000000000000000000000002' ], [ "ns1.example.org", 3600, 28, h'20010db8000000000000000000000035' ], [ "ns2.example.org", 3600, 28, h'20010db8000000000000000000003535' ] ] ]¶
This one advertises two local CoAP servers (identified by service name _coap._udp.local
) at
2001:db8::1 and 2001:db8::2 and two nameservers for the example.org domain, ns1.example.org at
2001:db8::35 and ns2.example.org at 2001.db8::3535. Each of the transmitted records has a TTL of
3600 seconds.¶
Fix Section 8.2.1 title¶
Amend for capability to carry more than one question¶
Hint at future of name compression in later draft versions¶
Use canonical name for CBOR-packed¶
Add Discussion section and note on compression¶
TODO acknowledge.¶
Carsten Bormann¶