title: DNS Queries over HTTPS abbrev: DNS Queries over HTTPS docname: draft-ietf-doh-dns-over-https
stand_alone: true
ipr: trust200902 area: Art kw: Internet-Draft cat: std
pi: toc: yes tocdepth: 4 sortrefs: yes symrefs: yes
ins: P. Hoffman name: Paul Hoffman org: ICANN email: paul.hoffman@icann.org
ins: P. McManus name: Patrick McManus org: Mozilla email: mcmanus@ducksong.com
normative: RFC1035: RFC2308: RFC4648: RFC5246: RFC6960: RFC7230: RFC7231: RFC7234: RFC7540: RFC7541: RFC8174:
informative: RFC2181: RFC5861: RFC6147: RFC6891: RFC6950: RFC6961: RFC7719: RFC7830: RFC7858: CORS: title: "Cross-Origin Resource Sharing" target: "https://fetch.spec.whatwg.org/#http-cors-protocol" QUERYPARAMETER: title: "application/x-www-form-urlencoded Parsing" target: "https://url.spec.whatwg.org/#application/x-www-form-urlencoded"
--- abstract
This document describes how to run DNS service over HTTP using https:// URIs.
[[ There is a repository for this draft at https://github.com/dohwg/draft-ietf-doh-dns-over-https ]].
--- middle
This document defines a specific protocol for sending DNS {{RFC1035}} queries and getting DNS responses over HTTP {{RFC7540}} using https:// (and therefore TLS {{RFC5246}} security for integrity and confidentiality). Each DNS query-response pair is mapped into a HTTP request-response pair.
The described approach is more than a tunnel over HTTP. It establishes default media formatting types for requests and responses but uses normal HTTP content negotiation mechanisms for selecting alternatives that endpoints may prefer in anticipation of serving new use cases. In addition to this media type negotiation, it aligns itself with HTTP features such as caching, redirection, proxying, authentication, and compression.
The integration with HTTP provides a transport suitable for both traditional DNS clients and native web applications seeking access to the DNS.
Two primary uses cases were considered during this protocol's development. They included preventing on-path devices from interfering with DNS operations and allowing web applications to access DNS information via existing browser APIs in a safe way consistent with Cross Origin Resource Sharing (CORS) {{CORS}}. There are certainly other uses for this work.
A server that supports this protocol is called a "DNS API server" to differentiate it from a "DNS server" (one that uses the regular DNS protocol). Similarly, a client that supports this protocol is called a "DNS API client".
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, RFC8174 {{RFC8174}} when, and only when, they appear in all capitals, as shown here.
The protocol described here bases its design on the following protocol requirements:
-
The protocol must use normal HTTP semantics.
-
The queries and responses must be able to be flexible enough to express every DNS query that would normally be sent in DNS over UDP (including queries and responses that use DNS extensions, but not those that require multiple responses).
-
The protocol must permit the addition of new formats for DNS queries and responses.
-
The protocol must ensure interoperability by specifying a single format for requests and responses that is mandatory to implement. That format must be able to support future modifications to the DNS protocol including the inclusion of one or more EDNS options (including those not yet defined).
-
The protocol must use a secure transport that meets the requirements for HTTPS.
-
Supporting network-specific DNS64 {{RFC6147}}
-
Supporting other network-specific inferences from plaintext DNS queries
-
Supporting insecure HTTP
To make a DNS API query a DNS API client encodes a single DNS query into an HTTP request using either the HTTP GET or POST method and the other requirements of this section. The DNS API server defines the URI used by the request. Configuration and discovery of the URI is done out of band from this protocol.
When using the POST method the DNS query is included as the message body of the HTTP request and the Content-Type request header indicates the media type of the message. POST-ed requests are smaller than their GET equivalents.
When using the GET method the URI path MUST contain a query parameter
name-value pair {{QUERYPARAMETER}}
with the name of "ct" and a value indicating the media-format used for
the dns parameter. The value may either be an explicit media type
(e.g., ct=application/dns-udpwireformat&dns=...) or it may be
empty. An empty value indicates the default
application/dns-udpwireformat type (e.g., ct&dns=...).
When using the GET method the URI path MUST contain a query parameter with the name of "dns". The value of the parameter is the content of the request potentially encoded with base64url {{RFC4648}}. Specifications that define media types for use with DOH, such as DNS Wire Format {{dnswire}} of this document, MUST indicate if the dns parameter uses base64url encoding.
Using the GET method is friendlier to many HTTP cache implementations.
The DNS API client SHOULD include an HTTP "Accept:" request header to say what type of content can be understood in response. The client MUST be prepared to process "application/dns-udpwireformat" {{dnswire}} responses but MAY process any other type it receives.
In order to maximize cache friendliness, DNS API clients using media formats that include DNS ID, such as application/dns-udpwireformat, SHOULD use a DNS ID of 0 in every DNS request. HTTP correlates request and response, thus eliminating the need for the ID in a media type such as application/dns-udpwireformat and the use of a varying DNS ID can cause semantically equivalent DNS queries to be cached separately.
DNS API clients can use HTTP/2 padding and compression in the same way that other HTTP/2 clients use (or don't use) them.
The data payload is the DNS on-the-wire format defined in {{RFC1035}}. The format is for DNS over UDP. Note that this is different than the wire format used in {{RFC7858}}. Also note that while {{RFC1035}} says "Messages carried by UDP are restricted to 512 bytes", that was later updated by {{RFC6891}}, and this protocol does not put any limit on the size of DNS on-the-wire format payloads.
When using the GET method, the data payload MUST be encoded with base64url {{RFC4648}} and then placed as a name value pair in the query portion of the URI with name "dns". Padding characters for base64url MUST NOT be included.
When using the POST method, the data payload MUST NOT be encoded and is used directly as the HTTP message body.
DNS API clients using the DNS wire format MAY have one or more EDNS options {{RFC6891}} in the request.
The media type is "application/dns-udpwireformat".
These examples use HTTP/2 style formatting from {{RFC7540}}.
These examples use a DNS API service located at https://dnsserver.example.net/dns-query to resolve the IN A records.
The requests are represented as application/dns-udpwirefomat typed bodies.
The first example request uses GET to request www.example.com
:method = GET
:scheme = https
:authority = dnsserver.example.net
:path = /dns-query?ct& (no space or CR)
dns=AAABAAABAAAAAAAAA3d3dwdleGFtcGxlA2NvbQAAAQAB
accept = application/dns-udpwireformat
The same DNS query for www.example.com, using the POST method would be:
:method = POST
:scheme = https
:authority = dnsserver.example.net
:path = /dns-query
accept = application/dns-udpwireformat
content-type = application/dns-udpwireformat
content-length = 33
<33 bytes represented by the following hex encoding>
00 00 01 00 00 01 00 00 00 00 00 00 03 77 77 77
07 65 78 61 6d 70 6c 65 03 63 6f 6d 00 00 01 00
01
Finally, a GET based query for a.62characterlabel-makes-base64url-distinct-from-standard-base64.example.com is shown as an example to emphasize that the encoding alphabet of base64url is different than regular base64 and that padding is omitted.
The DNS query is 94 bytes represented by the following hex encoding
00 00 01 00 00 01 00 00 00 00 00 00 01 61 3e 36
32 63 68 61 72 61 63 74 65 72 6c 61 62 65 6c 2d
6d 61 6b 65 73 2d 62 61 73 65 36 34 75 72 6c 2d
64 69 73 74 69 6e 63 74 2d 66 72 6f 6d 2d 73 74
61 6e 64 61 72 64 2d 62 61 73 65 36 34 07 65 78
61 6d 70 6c 65 03 63 6f 6d 00 00 01 00 01
:method = GET
:scheme = https
:authority = dnsserver.example.net
:path = /dns-query?ct& (no space or CR)
dns=AAABAAABAAAAAAAAAWE-NjJjaGFyYWN0ZXJsYWJl (no space or CR)
bC1tYWtlcy1iYXNlNjR1cmwtZGlzdGluY3QtZnJvbS1z (no space or CR)
dGFuZGFyZC1iYXNlNjQHZXhhbXBsZQNjb20AAAEAAQ
accept = application/dns-udpwireformat
An HTTP response with a 2xx status code ({{RFC7231}} Section 6.3) indicates a valid DNS response to the query made in the HTTP request. A valid DNS response includes both success and failure responses. For example, a DNS failure response such as SERVFAIL or NXDOMAIN will be the message in a successful 2xx HTTP response even though there was a failure at the DNS layer. Responses with non-successful HTTP status codes do not contain DNS answers to the question in the corresponding request. Some of these non-successful HTTP responses (e.g., redirects or authentication failures) could allow clients to make new requests to satisfy the original question.
Different response media types will provide more or less information from a DNS response. For example, one response type might include the information from the DNS header bytes while another might omit it. The amount and type of information that a media type gives is solely up to the format, and not defined in this protocol.
At the time this is published, the response types are works in progress. The only response type defined in this document is "application/dns-udpwireformat", but it is possible that other response formats will be defined in the future.
The DNS response for "application/dns-udpwireformat" in {{dnswire}} MAY have one or more EDNS options, depending on the extension definition of the extensions given in the DNS request.
Each DNS request-response pair is matched to one HTTP request-response pair. The responses may be processed and transported in any order using HTTP's multi-streaming functionality ({{RFC7540}} Section 5}).
The Answer section of a DNS response can contain zero or more RRsets. (RRsets are defined in {{RFC7719}}.) According to {{RFC2181}}, each resource record in an RRset has Time To Live (TTL) freshness information. Different RRsets in the Answer section can have different TTLs, although it is only possible for the HTTP response to have a single freshness lifetime. The HTTP response freshness lifetime ({{RFC7234}} Section 4.2) should be coordinated with the RRset with the smallest TTL in the Answer section of the response. Specifically, the HTTP freshness lifetime SHOULD be set to expire at the same time any of the DNS resource records in the Answer section reach a 0 TTL. The response freshness lifetime MUST NOT be greater than that indicated by the DNS resoruce record with the smallest TTL in the response.
If the DNS response has no records in the Answer section, and the DNS response has an SOA record in the Authority section, the response freshness lifetime MUST NOT be greater than the MINIMUM field from that SOA record. (See {{RFC2308}}.) Otherwise, the HTTP response MUST set a freshness lifetime ({{RFC7234}} Section 4.2) of 0 by using a mechanism such as "Cache-Control: no-cache" ({{RFC7234}} Section 5.2.1.4).
A DNS API client that receives a response without an explicit freshness lifetime MUST NOT assign that response a heuristic freshness ({{RFC7234}} Section 4.2.2.) greater than that indicated by the DNS Record with the smallest TTL in the response.
A DNS API server MUST be able to process application/dns-udpwireformat request messages.
A DNS API server SHOULD respond with HTTP status code 415 (Unsupported Media Type) upon receiving a media type it is unable to process.
This document does not change the definition of any HTTP response codes or otherwise proscribe their use.
This is an example response for a query for the IN A records for "www.example.com" with recursion turned on. The response bears one record with an address of 192.0.2.1 and a TTL of 128 seconds.
:status = 200
content-type = application/dns-udpwireformat
content-length = 64
cache-control = max-age=128
<64 bytes represented by the following hex encoding>
00 00 81 80 00 01 00 01 00 00 00 00 03 77 77 77
07 65 78 61 6d 70 6c 65 03 63 6f 6d 00 00 01 00
01 03 77 77 77 07 65 78 61 6d 70 6c 65 03 63 6f
6d 00 00 01 00 01 00 00 00 80 00 04 C0 00 02 01
This protocol MUST be used with the https scheme URI {{RFC7230}}.
A DNS API client may utilize a hierarchy of caches that include both HTTP and DNS specific caches. HTTP cache entries may be bypassed with HTTP mechanisms such as the "Cache-Control no-cache" directive; however DNS caches do not have a similar mechanism.
A DOH response that was previously stored in an HTTP cache will contain the {{RFC7234}} Age response header indicating the elapsed time between when the entry was placed in the HTTP cache and the current DOH response. DNS API clients should subtract this time from the DNS TTL if they are re-sharing the information in a non HTTP context (e.g., their own DNS cache) to determine the remaining time to live of the DNS record.
HTTP revalidation (e.g., via If-None-Match request headers) of cached DNS information may be of limited value to DOH as revalidation provides only a bandwidth benefit and DNS transactions are normally latency bound. Furthermore, the HTTP response headers that enable revalidation (such as "Last-Modified" and "Etag") are often fairly large when compared to the overall DNS response size, and have a variable nature that creates constant pressure on the HTTP/2 compression dictionary {{RFC7541}}. Other types of DNS data, such as zone transfers, may be larger and benefit more from revalidation. DNS API servers may wish to consider whether providing these validation enabling response headers is worthwhile.
The stale-while-revalidate and stale-if-error cache control directives may be well suited to a DOH implementation when allowed by server policy. Those mechanisms allow a client, at the server's discretion, to reuse a cache entry that is no longer fresh under some extenuating circumstances defined in {{RFC5861}}.
All HTTP servers, including DNS API servers, need to consider cache interaction when they generate responses that are not globally valid. For instance, if a DNS API server customized a response based on the client's identity then it would not want to globally allow reuse of that response. This could be accomplished through a variety of HTTP techniques such as a Cache-Control max-age of 0, or perhaps by the Vary response header.
The minimum version of HTTP used by DOH SHOULD be HTTP/2 {{RFC7540}}.
The messages in classic UDP based DNS {{RFC1035}} are inherently unordered and have low overhead. A competitive HTTP transport needs to support reordering, parallelism, priority, and header compression to achieve similar performance. Those features were introduced to HTTP in HTTP/2 {{RFC7540}}. Earlier versions of HTTP are capable of conveying the semantic requirements of DOH but may result in very poor performance for many uses cases.
Before using DOH response data for DNS resolution, the client MUST establish that the HTTP request URI is a trusted service for the DOH query. For HTTP requests initiated by the DNS API client this trust is implicit in the selection of URI. For HTTP server push ({{RFC7540}} Section 8.2) extra care must be taken to ensure that the pushed URI is one that the client would have directed the same query to if the client had initiated the request. This specification does not extend DNS resolution privileges to URIs that are not recognized by the client as trusted DNS API servers.
To: ietf-types@iana.org
Subject: Registration of MIME media type
application/dns-udpwireformat
MIME media type name: application
MIME subtype name: dns-udpwireformat
Required parameters: n/a
Optional parameters: n/a
Encoding considerations: This is a binary format. The contents are a
DNS message as defined in RFC 1035. The format used here is for DNS
over UDP, which is the format defined in the diagrams in RFC 1035.
Security considerations: The security considerations for carrying
this data are the same for carrying DNS without encryption.
Interoperability considerations: None.
Published specification: This document.
Applications that use this media type:
Systems that want to exchange full DNS messages.
Additional information:
Magic number(s): n/a
File extension(s): n/a
Macintosh file type code(s): n/a
Person & email address to contact for further information:
Paul Hoffman, paul.hoffman@icann.org
Intended usage: COMMON
Restrictions on usage: n/a
Author: Paul Hoffman, paul.hoffman@icann.org
Change controller: IESG
Running DNS over HTTPS relies on the security of the underlying HTTP transport. This mitigates classic amplification attacks for UDP-based DNS. Implementations utilizing HTTP/2 benefit from the TLS profile defined in {{RFC7540}} Section 9.2.
Session level encryption has well known weaknesses with respect to traffic analysis which might be particularly acute when dealing with DNS queries. HTTP/2 provides further advice about the use of compression (Section 10.6 of {{RFC7540}}) and padding (Section 10.7 of {{RFC7540}}).
The HTTPS connection provides transport security for the interaction between the DNS API server and client, but does not inherently ensure the authenticity of DNS data. A DNS API client may also perform full DNSSEC validation of answers received from a DNS API server or it may choose to trust answers from a particular DNS API server, much as a DNS client might choose to trust answers from its recursive DNS resolver. This capability might be affected by the response media type.
{{caching}} describes the interaction of this protocol with HTTP caching. An adversary that can control the cache used by the client can affect that client's view of the DNS. This is no different than the security implications of HTTP caching for other protocols that use HTTP.
A server that is acting both as a normal web server and a DNS API server is in a position to choose which DNS names it forces a client to resolve (through its web service) and also be the one to answer those queries (through its DNS API service). An untrusted DNS API server can thus easily cause damage by poisoning a client's cache with names that the DNS API server chooses to poison. A client MUST NOT trust a DNS API server simply because it was discovered, or because the client was told to trust the DNS API server by an untrusted party. Instead, a client MUST only trust DNS API server that is configured as trustworthy.
A client can use DNS over HTTPS as one of multiple mechanisms to obtain DNS data. If a client of this protocol encounters an HTTP error after sending a DNS query, and then falls back to a different DNS retrieval mechanism, doing so can weaken the privacy expected by the user of the client.
Local policy considerations and similar factors mean different DNS servers may provide different results to the same query: for instance in split DNS configurations {{RFC6950}}. It logically follows that the server which is queried can influence the end result. Therefore a client’s choice of DNS server may affect the responses it gets to its queries. For example, in the case of DNS64 {{RFC6147}}, the choice could affect whether IPv6/IPv4 translation will work at all.
The HTTPS channel used by this specification establishes secure two party communication between the DNS API client and the DNS API server. Filtering or inspection systems that rely on unsecured transport of DNS will not function in a DNS over HTTPS environment.
Some HTTPS client implementations perform real time third party checks of the revocation status of the certificates being used by TLS. If this check is done as part of the DNS API server connection procedure and the check itself requires DNS resolution to connect to the third party a deadlock can occur. The use of an OCSP {{RFC6960}} server is one example of how this deadlock can happen. To avoid the possibility of deadlock, DNS API servers SHOULD bundle the certificate status as part of the handshake using a mechanism appropriate to the version of TLS, such as using {{RFC6961}} for TLS version 1.2.
A DNS API client may face a similar bootstrapping problem when the HTTP request needs to resolve the hostname portion of the DNS URI. Just as the address of a traditional DNS nameserver cannot be originally determined from that same server, a DNS API client cannot use its DNS API server to initially resolve the server's host name into an address. Alternative strategies a client might employ include making the initial resolution part of the configuration, IP based URIs and corresponding IP based certificates for HTTPS, or resolving the DNS API server's hostname via traditional DNS or another DNS API server while still authenticating the resulting connection via HTTPS.
HTTP {{RFC7230}} is a stateless application level protocol and therefore DOH implementations do not provide stateful ordering guarantees between different requests. DOH cannot be used as a transport for other protocols that require strict ordering.
If a DNS API server responds to a DNS API client with a DNS message that has the TC (truncation) bit set in the header, that indicates that the DNS API server was not able to retrieve a full answer for the query over TCP and is providing the best answer it could get. This protocol does not require that a DNS API server that cannot get an untruncated answer send back such an answer; it can instead send back an HTTP error to indicate that it cannot give a useful answer.
Joe Hildebrand contributed lots of material for a different iteration of this document. Helpful early comments were given by Ben Schwartz and Mark Nottingham.
--- back
The following is an incomplete list of earlier work that related to DNS over HTTP/1 or representing DNS data in other formats.
The list includes links to the tools.ietf.org site (because these documents are all expired) and web sites of software.