Internet DRAFT - draft-dkg-dprive-demux-dns-http

draft-dkg-dprive-demux-dns-http







dprive                                                        D. Gillmor
Internet-Draft                                                      ACLU
Updates: 1035, 7230 (if approved)                           May 17, 2017
Intended status: Informational
Expires: November 18, 2017


               Demultiplexing Streamed DNS from HTTP/1.x
                   draft-dkg-dprive-demux-dns-http-03

Abstract

   DNS over TCP and HTTP/1.x are both stream-oriented, client-speaks-
   first protocols.  They can both be run over a stream-based security
   protocol like TLS.  A server accepting a stream-based client can
   distinguish between a valid stream of DNS queries and valid stream of
   HTTP/1.x requests by simple observation of the first few octets sent
   by the client.  This can be done without any external demultiplexing
   mechanism like TCP port number or ALPN.

   Implicit multiplexing of the two protocols over a single listening
   port can be useful for obscuring the presence of DNS queries from a
   network observer, which makes it relevant for DNS privacy.

   Widespread adoption of the described approach could constrain
   evolution of the stream-based variants of both DNS ([RFC1035]) and
   HTTP/1.x ([RFC7230]) by ossifying existing distinguishing bit
   patterns in early octets sent by the client.  However, this draft
   explicitly rules out multiplexing in this form with HTTP/2, so it
   should place no constraints on it or any higher version of HTTP.

Status of This Memo

   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 http://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 November 18, 2017.




Gillmor                 Expires November 18, 2017               [Page 1]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


Copyright Notice

   Copyright (c) 2017 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
   (http://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 Simplified BSD License text as described in Section 4.e of
   the Trust Legal Provisions and are provided without warranty as
   described in the Simplified BSD License.

Table of Contents

   1.  Introduction  . . . . . . . . . . . . . . . . . . . . . . . .   3
     1.1.  Terminology . . . . . . . . . . . . . . . . . . . . . . .   4
   2.  Scoping . . . . . . . . . . . . . . . . . . . . . . . . . . .   4
     2.1.  Distinguish only at the start of a stream . . . . . . . .   4
     2.2.  HTTP/2 is not always client-speaks-first  . . . . . . . .   4
     2.3.  Avoid multiplexing in the clear . . . . . . . . . . . . .   5
     2.4.  Avoid mixing with other demultiplexing  . . . . . . . . .   5
     2.5.  Heavily-restricted network environments . . . . . . . . .   5
     2.6.  Why not ALPN? . . . . . . . . . . . . . . . . . . . . . .   5
   3.  Overview of initial octets  . . . . . . . . . . . . . . . . .   6
     3.1.  DNS stream initial octets . . . . . . . . . . . . . . . .   6
     3.2.  HTTP/1.x initial octets . . . . . . . . . . . . . . . . .   7
       3.2.1.  HTTP/0.9  . . . . . . . . . . . . . . . . . . . . . .   7
       3.2.2.  HTTP/1.0 and HTTP/1.1 . . . . . . . . . . . . . . . .   8
   4.  Specific octets . . . . . . . . . . . . . . . . . . . . . . .   9
     4.1.  octets 0 and 1  . . . . . . . . . . . . . . . . . . . . .   9
     4.2.  octets 2 and 3  . . . . . . . . . . . . . . . . . . . . .   9
     4.3.  octet 4 . . . . . . . . . . . . . . . . . . . . . . . . .  10
     4.4.  octet 5 . . . . . . . . . . . . . . . . . . . . . . . . .  10
     4.5.  octets 6 and 7  . . . . . . . . . . . . . . . . . . . . .  11
     4.6.  octets 8 through 11 . . . . . . . . . . . . . . . . . . .  11
     4.7.  octets 12 and 13  . . . . . . . . . . . . . . . . . . . .  11
   5.  Combinations of octets  . . . . . . . . . . . . . . . . . . .  11
     5.1.  Proof: a valid DNS message cannot be an HTTP/1.x query  .  12
   6.  Guidance for Demultiplexing Servers . . . . . . . . . . . . .  13
     6.1.  Without supporting HTTP/0.9 . . . . . . . . . . . . . . .  13
     6.2.  Supporting archaic HTTP/0.9 clients . . . . . . . . . . .  13
     6.3.  Signaling demultiplexing capacity . . . . . . . . . . . .  14
   7.  Guidance for DNS clients  . . . . . . . . . . . . . . . . . .  15
     7.1.  Interpreting failure  . . . . . . . . . . . . . . . . . .  16
   8.  Guidance for HTTP clients . . . . . . . . . . . . . . . . . .  16



Gillmor                 Expires November 18, 2017               [Page 2]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


   9.  Security Considerations . . . . . . . . . . . . . . . . . . .  16
   10. Privacy Considerations  . . . . . . . . . . . . . . . . . . .  16
   11. IANA Considerations . . . . . . . . . . . . . . . . . . . . .  17
   12. Document Considerations . . . . . . . . . . . . . . . . . . .  17
   13. References  . . . . . . . . . . . . . . . . . . . . . . . . .  17
     13.1.  Normative References . . . . . . . . . . . . . . . . . .  17
     13.2.  Informative References . . . . . . . . . . . . . . . . .  18
   Author's Address  . . . . . . . . . . . . . . . . . . . . . . . .  19

1.  Introduction

   DNS and HTTP/1.x are both client-speaks-first protocols capable of
   running over stream-based transport like TCP, or as the payload of a
   typical TLS [RFC5246] session.

   There are some contexts where it is useful for a server to be able to
   decide what protocol is used by an incoming TCP stream, to choose
   dynamically between DNS and HTTP/1.x on the basis of the stream
   itself (rather than a port designation or other explicit
   demultiplexing).

   For example, a TLS terminator listening on port 443 and receiving
   either no ALPN token at all, or the "http/1.1" ALPN token might be
   willing to serve DNS-over-TLS [RFC7858] as well as HTTPS.

   A simple demultiplexing server should do this demuxing based on the
   first few bytes sent by the client on a given stream; once a choice
   has been established, the rest of the stream is committed to one or
   the other interpretation.

   This document provides proof that a demultiplexer can robustly
   distinguish HTTP/1.x from DNS on the basis of the content of the
   first few bytes of the client's stream alone.

   A DNS client that knows it is talking to a server which is this
   position (e.g. trying to do DNS-over-TLS on TCP port 443 with no ALPN
   token, used traditionally only for HTTPS) might also want to be aware
   of network traffic patterns that could confuse such a server.  This
   document presents explicit mitigations that such a DNS client MAY
   decide to use.

   This document limits its discussion to HTTP/1.x over TCP or TLS or
   some other classical stream-based protocol (it excludes HTTP over
   QUIC, for example, and HTTP/2 [RFC7540] or later).  Likewise, it
   considers only the TCP variant of DNS (and excludes DNS over UDP or
   any other datagram transport).





Gillmor                 Expires November 18, 2017               [Page 3]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


1.1.  Terminology

   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
   [RFC2119].

2.  Scoping

2.1.  Distinguish only at the start of a stream

   A server which attempts to distinguish DNS queries from HTTP/1.x
   requests individually might consider using these guidelines in the
   middle of a running stream (e.g. at natural boundaries, like the end
   of an HTTP/1.1 request, or after a DNS message), but this document
   focuses specifically on a heuristic choice for the whole stream,
   based on the initial few octets sent by the client.

   While it's tempting to consider distinguishing at multiple points in
   the stream, the complexities of determining the specific end of an
   HTTP/1.x request body and handling HTTP/1.x error cases make this
   more difficult to implement on the side of a DNS client configured to
   talk to such a server.  Interleaving the responses themselves on a
   stream with multiple data elements is also challenging.  So do not
   use this technique anywhere but at the beginning of a stream!

   If being able to interleave DNS queries with HTTP requests on a
   single stream is desired, a strategy like
   [I-D.hoffman-dns-over-https] or [I-D.ietf-dnsop-dns-wireformat-http]
   is recommended instead.

2.2.  HTTP/2 is not always client-speaks-first

   While this demultiplexing technique functions for HTTP/1.0 and
   HTTP/1.1, it does not work for HTTP/2 [RFC7540] because HTTP/2 is not
   guaranteed to be a client-speaks-first protocol.  In particular, many
   HTTP/2 servers prefer to send a SETTINGS frame immediately without
   waiting for data from the client, if they already know they're
   speaking HTTP/2.  In the event that HTTP/2 is to be transported over
   TLS, the ALPN token negotiated in the TLS handshake is "h2", which
   allows the server to know as soon as the handshake is complete that
   it can start pushing data to the client.

   A standard DNS-over-TLS client connecting to a server that might be
   multiplexing DNS with HTTP on the same listener MUST NOT indicate an
   intent to speak HTTP/2 that could prompt this unsolicited first
   flight from the server.  Concretely, a DNS client connecting over TLS




Gillmor                 Expires November 18, 2017               [Page 4]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


   on TCP port 443 expecting to speak standard DNS-over-TLS [RFC7858]
   MUST NOT offer or accept the "h2" ALPN token.

   If use of DNS in the same channel as HTTP/2 is deisred, a strategy
   like [I-D.hoffman-dns-over-https] is recommended instead.

2.3.  Avoid multiplexing in the clear

   The widespread deployment of transparent HTTP/1.x proxies makes it
   likely that any attempt to do this kind of multiplexing/
   demultiplexing on a cleartext channel that normally carries HTTP/1.x
   (e.g.  TCP port 80) will fail or trigger other "interesting"
   behaviors.  The approach described in this draft should be done only
   in channels sufficiently obscured that a transparent proxy would not
   try to interpret the resultant stream.

2.4.  Avoid mixing with other demultiplexing

   Some other (non-IETF) systems (e.g.  [HAPROXY]) take a similar
   approach with multiplexing data on top of HTTP/1.x by taking
   advantage of bitpatterns that are presumed to not be present in
   normal HTTP/1.x requests.

   Use of the approach described in this draft in conjunction with these
   other approaches is not advisable.  Doing so safely would require
   explicit and detailed review of all three (or more) protocols
   involved.

2.5.  Heavily-restricted network environments

   Some network environments are so tightly constrained that outbound
   connections on standard TCP ports are not accessible.  In some of
   these environments, an explicit HTTP proxy is available, and clients
   must use the HTTP CONNECT pseudo-method to make https connections.
   While this multiplexing approach can be used in such a restrictive
   environment, it would be necessary to teach the DNS client how to
   talk to (and through) the HTTP proxy.  These details are out of scope
   for this document.  A DNS client capable of this additional layer of
   complexity may prefer to pursue a strategy like
   [I-D.hoffman-dns-over-https] instead.

2.6.  Why not ALPN?

   If this is done over TLS, a natural question is whether the client
   should simply indicate its preferred protocol in the TLS handshake's
   ALPN [RFC7301] extension (e.g. with some new ALPN token "dns").





Gillmor                 Expires November 18, 2017               [Page 5]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


   However, ALPN tokens requested by the client are visible to a network
   observer (and the ALPN token selected by the server is visible to a
   network observer in TLS 1.2 and earlier), so a network controller
   attempting to confine the user's DNS traffic to a limited set of
   servers could use the ALPN extension as a signal to block DNS-
   specific streams.

   Another alternative could be an ALPN token that indicates
   potentially-multiplexed traffic (e.g. "http/1.1-or-dns").  This has a
   comparable problem when confronted with a network adversary that
   intends to penalize or hamper DNS-over-TLS.  Existing HTTP clients
   will not send this token, and even if some start to offer it, it will
   provide less cover for DNS-over-TLS clients.

3.  Overview of initial octets

3.1.  DNS stream initial octets

   [RFC1035] section 4.2.2 ("TCP Usage") shows that every stream-based
   DNS connection starts with a DNS message, preceded with a 2-octet
   message length field:

   The message is prefixed with a two byte length field which gives
   the message length, excluding the two byte length field.

   [RFC6895] section 2 represents the DNS message header section, which
   is the first part of the DNS message on the wire (after the message
   length).

                                  1  1  1  1  1  1
    0  1  2  3  4  5  6  7  8  9  0  1  2  3  4  5
   +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
   |                      ID                       |
   +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
   |QR|   OpCode  |AA|TC|RD|RA| Z|AD|CD|   RCODE   |
   +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
   |                QDCOUNT/ZOCOUNT                |
   +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
   |                ANCOUNT/PRCOUNT                |
   +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
   |                NSCOUNT/UPCOUNT                |
   +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
   |                    ARCOUNT                    |
   +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+

   So in a DNS over TCP stream, the interpretation of the initial 14
   octets are fixed based on information about the first query sent on
   the stream:



Gillmor                 Expires November 18, 2017               [Page 6]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


   o  0,1: length of initial DNS message

   o  2,3: DNS Transaction ID

   o  4,5: DNS opcode, flags, and response code

   o  6,7: Question count (or Zone count in UPDATE)

   o  8,9: Answer count (or Prerequisite count in UPDATE)

   o  10,11: Authority count (or Update count in UPDATE)

   o  12,13: Additional RR count

   All DNS streams sent over TCP start with at least these 14 octets.

3.2.  HTTP/1.x initial octets

   In an HTTP stream before HTTP/2, the first octets sent from the
   client are either the so-called "Simple-Request" (for HTTP/0.9) or
   the "Request-Line" (for HTTP/1.0 and HTTP/1.1).  The data in this
   initial stream has variable characteristics.

   Most servers may wish to ignore the oldest of these, HTTP/0.9.

3.2.1.  HTTP/0.9

   [RFC1945] section 4.1 says that HTTP/0.9 queries (that is, HTTP
   queries from before HTTP/1.0 was formalized) use this form:

   Simple-Request = "GET" SP Request-URI CRLF

   Note that HTTP/0.9 clients send this string and only this string,
   nothing else (no request body, no subsequent requests).  The
   "Request-URI" token is guaranteed to start with a printable ASCII
   character, and cannot contain any members of the CTL class (values
   0x00 through 0x1F) but due to loose early specifications, it might
   sometimes contain high-valued octets (those with the most-significant
   bit set - 0x80 or above).

   So the first 5 octets are all constrained to be no less than 0x20
   (SP) and no more than 0x7F (DEL), and all subsequent octets sent from
   the client have a value at least 0x0A (LF).

   The shortest possible HTTP/0.9 client request is:

    char: G  E  T  SP /  CR LF
   index: 0  1  2  3  4  5  6



Gillmor                 Expires November 18, 2017               [Page 7]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


   The lowest possible HTTP/0.9 client request (sorted ASCIIbetically)
   is:

    char: G  E  T  SP +  :  CR LF
   index: 0  1  2  3  4  5  6  7

3.2.2.  HTTP/1.0 and HTTP/1.1

   The request line format for HTTP/1.1 matches that of HTTP/1.0
   (HTTP/1.1 adds protocol features like pipelining, but doesn't change
   the request form itself).  But unlike HTTP/0.9, the initial verb (the
   "method") can vary.

   [RFC7230] section 3.1.1 says that the first line of an HTTP/1.1
   request is:

   request-line   = method SP request-target SP HTTP-version CRLF
   method         = token

   and [RFC7230] section 3.2.6 says:

    token          = 1*tchar

    tchar          = "!" / "#" / "$" / "%" / "&" / "'" / "*"
                   / "+" / "-" / "." / "^" / "_" / "`" / "|" / "~"
                   / DIGIT / ALPHA
                   ; any VCHAR, except delimiters

   and VCHAR is defined in [RFC5234] appendix B.1 as:

   VCHAR          =  %x21-7E

   "request-target" itself cannot contain 0x20 (SP) or any CTL
   characters, or any characters above the US-ASCII range (> 0x7F).

   And the "HTTP-version" token is either the literal string "HTTP/1.0"
   or the literal string "HTTP/1.1", both of which are constrained to
   the same printable-ASCII range.

   The ASCIIbetically-lowest shortest possible HTTP/1.0 or HTTP/1.1
   request is:

    char: !  SP /  SP H  T  T  P  /  1  .  0  CR LF CR LF
   index: 0  1  2  3  4  5  6  7  8  9  0  a  b  c  d  e

   In any case, no HTTP/1.0 or HTTP/1.1 request line can include any
   values lower than 0x0A (LF) or greater than 0x7F (DEL) in the first
   15 octets.



Gillmor                 Expires November 18, 2017               [Page 8]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


   However, [RFC7230] section 3.1.1 also says:

   In the interest of robustness, a server that is expecting to receive
   and parse a request-line SHOULD ignore at least one empty line (CRLF)
   received prior to the request-line.

   So we should also consider accepting an arbitrary number of repeated
   CRLF sequences before the request-line as a potentially-valid HTTP
   client behavior.

4.  Specific octets

   The sections below examine likely values of specific octet positions
   in the stream.  All octet indexes are 0-based.

4.1.  octets 0 and 1

   Any DNS message less than 3338 octets sent as the initial query over
   TCP can be reliably distinguished from any version of HTTP/1.x by the
   first two octets of the TCP stream alone.

   3338 is 0x0D0A, or the ASCII string CRLF, which some HTTP/1.x clients
   might send before an initial request.  No HTTP/1.x client can
   legitimately send anything lower than this.

   Most DNS queries are easily within this range automatically.

4.2.  octets 2 and 3

   In a DNS stream, octets 2 and 3 represent the client-chosen message
   ID.  The message ID is used to bind messages with responses.  Over
   connectionless transports like UDP, this is an important anti-
   spoofing measure, as well as a distinguishing measure for clients
   reusing the same UDP port for multiple outstanding queries.  Standard
   DNS clients already explicitly randomize this value.

   For the connection-oriented streaming DNS discussed here, the anti-
   spoofing characteristics are not relevant (the connection itself
   provides anti-spoofing), so the client is free to choose arbitrary
   values.

   With a standard DNS client which fully-randomizes these values, only
   25% of generated queries will have the high bits of both octets set
   to 0.  100% of all HTTP/1.x requests will have the high bits of both
   of these octets cleared.  Similarly, some small percentage of
   randomly-generated DNS queries will have values here lower than 0x0A,
   while no HTTP/1.x clients will ever send these low values.




Gillmor                 Expires November 18, 2017               [Page 9]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


4.3.  octet 4

   In a DNS stream, octet 4 combines several fields:

     0  1  2  3  4  5  6  7
   +--+--+--+--+--+--+--+--+
   |QR|   Opcode  |AA|TC|RD|
   +--+--+--+--+--+--+--+--+

   In a standard DNS query sent over a streaming interface, QR, Opcode,
   AA, and TC are all set to 0.  The least-significant bit (RD -
   Recursion Desired) is set when a packet is sent from a stub to a
   recursive resolver.  The value of such an octet is 0x01.  This value
   never occurs in octet 4 of a legitimate HTTP/1.x client.

   But under DNS UPDATE ([RFC2136], Opcode is set to 5 and all the
   option bits are cleared, which means this value would have 0x40
   (ASCII '@'), which could legitimately occur in some HTTP/1.x requests
   at this position.

4.4.  octet 5

   In a DNS stream, octet 5 also combines several fields:

     0  1  2  3  4  5  6  7
   +--+--+--+--+--+--+--+--+
   |RA| Z|AD|CD|   RCODE   |
   +--+--+--+--+--+--+--+--+

   In some DNS messages sent from a client, all these bits are 0.
   However, section 5.7 of [RFC6840] suggests that queries may wish to
   set the AD bit to indicate a desire to learn from a validating
   resolver whether the resolver considers the contents to be Authentic
   Data.

   [RFC6840] also suggests that:

   validating resolvers SHOULD set the CD bit on every upstream query.

   So many queries, particularly from DNSSEC-validating DNS clients, are
   likely to set bits 2 and 3, resulting in a value 0x30 (ASCII '0').
   This is usually a legitimate value for octet 5 in an HTTP/1.x
   request.








Gillmor                 Expires November 18, 2017              [Page 10]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


4.5.  octets 6 and 7

   In DNS, octets 6 and 7 represent the query count.  Most DNS clients
   will send one query at a time, which makes this value 0x0001.  As
   long as the number of initial queries does not exceed 0x0A0A (2570),
   then at least one of these octets will have a value less than 0x0A.
   No HTTP/1.x client sends an octet less than 0x0A in positions 6 or 7.

   In DNS UPDATE, octets 6 and 7 represent the zone count.  Entries in
   the Zone section of the DNS UPDATE message are structured identically
   to entries in the Query section of a standard DNS message.

4.6.  octets 8 through 11

   In streaming DNS, octets 8 through 11 represent answer counts and
   authority counts in normal DNS queries, or Prerequisite and Update
   counts in DNS UPDATE.  Standard DNS queries will set them both 0.
   DNS UPDATE queries are likely to include some records in these
   sections, so they won't be all zero, but as long as no more than 2570
   Prerequisite records and no more than 2570 Update records are sent,
   at least one octet will have value less than 0x0A.  But no HTTP/1.x
   client sends an octet less than 0x0A in these positions.

4.7.  octets 12 and 13

   In streaming DNS, octets 12 and 13 represent the number of Additional
   RRs.  When a DNS query is sent with EDNS(0), the OPT RR is accounted
   for here.  So this is often either 0x0000 or 0x0001.  In a Secure DNS
   UPDATE [RFC3007], the SIG(0) or TSIG record is also found in this
   section, which could increase the values of these octets to 0x0002.
   No HTTP/1.x client will send octets with these low values at these
   positions.

5.  Combinations of octets

   In a DNS message, each Question in the Question section (or Zone in
   the Zone section for DNS UPDATE) is at least 5 octets (1 octet for
   zero-length QNAME + 2 octets for QTYPE + 2 octets for QCLASS), and
   each RR (in the Answer, Authority, and Additional sections for normal
   DNS queries; or in the Prerequisite, Update, and Additional sections
   for DNS UPDATE) is at least 11 octets.  And the header itself is 12
   octets.

   So we know that for a valid DNS stream, the first message has a size
   of at least:






Gillmor                 Expires November 18, 2017              [Page 11]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


   min_first_msg_size = 12 + 5 * (256*o[6] + o[7]) +
                        11 * (256*(o[8] + o[10] + o[12]) +
                                   o[9] + o[11] + o[13])

   It's possible to compare this value with the expected first query
   size:

   first_msg_size = 256 * o[0] + o[1]

   if "first_query_size" is less than "min_first_query_size" we can be
   confident that the stream is not DNS.

5.1.  Proof: a valid DNS message cannot be an HTTP/1.x query

   For any a valid, stream-based DNS message:

   o  If there are fewer than 0x0A00 Questions then octet 6 < 0x0A.

   o  If there are fewer than 0x0A00 Answer RRs, then octet 8 < 0x0A.

   o  If there are fewer than 0x0A00 Authority RRs, then octet 10 <
      0x0A.

   o  If there are fewer than 0x0A00 Additional RRs, then octet 12 <
      0x0A.

   If any of these four inequalities hold, then the packet is clearly
   DNS, not HTTP/1.x.

   if none of them hold, then there are at least 0x0A00 (2560) Questions
   and 3*2560 == 7680 RRs.  But:

   12 + 5*2560 + 11*7680 == 97292

   So the smallest possible DNS message where none of these four
   inequalities hold is 97292 octets.  But a DNS message is limited in
   size to 65535 octets.

   Therefore at least one of these inequalities holds, and one of the
   first 14 octets of a DNS steam is < 0x0A.

   But in a standard HTTP/1.x request, none of the first 14 octets can
   have a value < 0x0A, so a valid DNS message cannot be mistaken for an
   HTTP/1.x request.







Gillmor                 Expires November 18, 2017              [Page 12]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


6.  Guidance for Demultiplexing Servers

   Upon receiving a connection stream that might be either DNS or
   HTTP/1.x, a server can inspect the initial octets of the stream to
   decide where to send it.

6.1.  Without supporting HTTP/0.9

   A server that doesn't care about HTTP/0.9 can simply wait for the
   first 14 octets of the client's request to come in.  Then the
   algorithm is:

   bytestream = read_from_client(14)
   for x in bytestream:
       if (x < 0x0A) or (x > 0x7F):
          return `DNS`
   return `HTTP`

6.2.  Supporting archaic HTTP/0.9 clients

   A server that decides to try to support HTTP/0.9 clients has a
   slightly more challenging task, since some of them may send fewer
   octets than the initial DNS message, and the server shouldn't block
   waiting for data that will never come.



























Gillmor                 Expires November 18, 2017              [Page 13]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


   bytestream = read_from_client(5)
   for x in bytestream[0:5]
       if (x < 0x0A) or (x > 0x7F):
          return `DNS`
   if (bytestream[0:4] != 'GET '):    # not HTTP/0.9
       bytestream += read_from_client(9)
       for x in bytestream[5:14]:
           if (x < 0x0A) or (x > 0x7f):
               return `DNS`
       return `HTTP`
   else:                              # maybe HTTP/0.9
       seen_sp = False
       seen_high = False
       while (len(bytestream) < 14):
           if (seen_sp and seen_high):
               return `DNS`
           x = read_from_client(1)
           bytestream += x
           if (x > 0x7F):
               seen_high = True
           elif (x < 0x0A):
               return `DNS`
           elif (x == 0x20):
               seen_sp = True       # SP found before CRLF, not HTTP/0.9
           elif (x == 0x0A):
               return `HTTP`
       return `HTTP`

   Note that if read_from_client() ever fails to read the number of
   requested bytes (e.g. because of EOF), then the stream is neither
   valid HTTP nor valid DNS, and can be discarded.

6.3.  Signaling demultiplexing capacity

   This document assumes that clients can learn out-of-band which
   listening service they can connect to.  For example, the
   administrator of a machine can configure a local forwarding stub
   resolver to use DNS-over-TLS on port 443 of some specific server.
   This explicit configuration carries with it some level of trust - the
   client is choosing to trust the configured server with its DNS
   queries.

   In some circumstances, it might be useful for a listener to signal to
   a client that it is willing and capable of handling both DNS and
   HTTP/1.x traffic.  While such signalling could be useful for dynamic
   discovery, it opens questions of trust (which servers should the
   client be willing to rely on for DNS resolution?) and is out-of-scope
   for this draft.



Gillmor                 Expires November 18, 2017              [Page 14]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


7.  Guidance for DNS clients

   Consider a DNS client that connects to a server that might be
   interested in answering HTTP/1.x requests on the same address/port
   (or other channel identifier).  The client wants to send traffic that
   is unambiguously DNS traffic to make it easy for the server to
   distinguish it from inbound HTTP/1.x requests.  Fortunately, this is
   trivial to do.  In fact, any sensibly-implemented DNS-over-TLS client
   can use this approach without modification, just by adjusting the
   port number of the upstream recursive resolver from 853 to 443.

   Such a client should follow these guidelines:

   o  Send the DNS message size (a 16-bit integer) together in the same
      packet with the full header of the first DNS message so that the
      recipient can review as much as possible of the frame at once.
      This is a best practice for efficient stream-based DNS anyway.

   If the client is concerned about stream fragmentation that it cannot
   control, and it is talking to a server that might be expecting
   HTTP/0.9 clients, then the server might not be willing to wait for
   the full initial 14 octets to make a decision.

   Note that this fragmentation is not a concern for streams wrapped in
   TLS when using modern AEAD ciphersuites.  In this case, the client
   gets to choose the size of the plaintext record, which is either
   recovered by the server in full (unfragmented) or the connection
   fails.

   If the client does not have such a guarantee from the transport, it
   MAY also take one of the following mitigating actions relating to the
   first DNS message it sends in the stream [explanation of what the
   server gets to see in the fragmented stream case are in square
   brackets after each mitigation]:

   o  Ensure the first message is marked as a query (QR = 0), and it
      uses opcode 0 ("Standard Query").  [bytestream[4] < 0x08]

   o  Ensure that the first message has RA = 0, Z = 0, and RCODE = 0.
      [bytestream[5] == 0x00]

   o  Ensure that the high bit of the first octet of the message ID of
      the first message is set. [bytestream[2] > 0x7F]

   o  Send an initial short Server Status DNS message ahead of the
      otherwise intended initial DNS message. [bytestream[0] == 0x00]





Gillmor                 Expires November 18, 2017              [Page 15]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


   o  Use the EDNS(0) padding option [RFC7830] to pad the first message
      to a multiple of 256 octets. [bytestream[1] == 0x00]

7.1.  Interpreting failure

   FIXME: A DNS client that does not already know that a server is
   willing to carry both types of traffic SHOULD expect a transport
   connection failure of some sort.  Can we say something specific about
   what it should expect?

8.  Guidance for HTTP clients

   HTTP clients SHOULD NOT send HTTP/0.9 requests, since modern HTTP
   servers are not required to support HTTP/0.9.  Sending an HTTP/1.0
   request (or any later version) is sufficient for a server to be able
   to distinguish the two protocols.

9.  Security Considerations

   FIXME: Clients should locally validate DNSSEC (servers may still be
   able to omit some records)

   FIXME: if widely deployed, consider amplification for DDoS against
   authoritative servers?

   FIXME: consider DNSSEC transparency

   FIXME: consider TLS session resumption - this counts as a new stream
   boundary, so the multiplexing decision need not persist across
   resumption.

   FIXME: consider 0-RTT

   FIXME: consider X.509 cert validation

   FIXME: what other security considerations should clients take?

   FIXME: what other security considerations should servers take?

10.  Privacy Considerations

   FIXME: DNS queries and HTTP requests can reveal potentially sensitive
   information about the sender.

   FIXME: consider DNS and HTTP traffic analysis - how should requests
   or responses be padded, aggregated, or delayed given that streams are
   multiplexed?




Gillmor                 Expires November 18, 2017              [Page 16]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


   FIXME: any other privacy considerations?

11.  IANA Considerations

   This document does not ask IANA to make any changes to existing
   registries.

   However, it does update the DNS and HTTP specifications, to reflect
   the fact that services using this demultiplexing technique may be
   constrained in adoption of future versions of either stream-based DNS
   or HTTP/1.x if those future versions modify either protocol in a way
   that breaks with the distinctions documented here.

   In particular, this draft assumes that all future stream-based
   versions of HTTP/1.x should have the following properties:

   o  the client will speak first

   o  the client will send at least 14 octets before expecting a
      response from the server.

   o  none of those first 14 octets will be below 0x0A (LF) or above
      0x7F (DEL).

   Future extensions to stream-based DNS or HTTP/1.x should take this
   demultiplexing technique into consideration.

12.  Document Considerations

   [ RFC Editor: please remove this section before publication ]

   This document is currently edited as markdown.  Minor editorial
   changes can be suggested via merge requests at
   https://gitlab.com/dkg/hddemux or by e-mail to the author.  Please
   direct all significant commentary to the public IETF DNS Privacy
   mailing list: dns-privacy@ietf.org or to the IETF HTTP WG mailing
   list: ietf-http-wg@w3.org

13.  References

13.1.  Normative References

   [RFC1035]  Mockapetris, P., "Domain names - implementation and
              specification", STD 13, RFC 1035, DOI 10.17487/RFC1035,
              November 1987, <http://www.rfc-editor.org/info/rfc1035>.






Gillmor                 Expires November 18, 2017              [Page 17]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


   [RFC1945]  Berners-Lee, T., Fielding, R., and H. Frystyk, "Hypertext
              Transfer Protocol -- HTTP/1.0", RFC 1945,
              DOI 10.17487/RFC1945, May 1996,
              <http://www.rfc-editor.org/info/rfc1945>.

   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119,
              DOI 10.17487/RFC2119, March 1997,
              <http://www.rfc-editor.org/info/rfc2119>.

   [RFC2136]  Vixie, P., Ed., Thomson, S., Rekhter, Y., and J. Bound,
              "Dynamic Updates in the Domain Name System (DNS UPDATE)",
              RFC 2136, DOI 10.17487/RFC2136, April 1997,
              <http://www.rfc-editor.org/info/rfc2136>.

   [RFC5234]  Crocker, D., Ed. and P. Overell, "Augmented BNF for Syntax
              Specifications: ABNF", STD 68, RFC 5234,
              DOI 10.17487/RFC5234, January 2008,
              <http://www.rfc-editor.org/info/rfc5234>.

   [RFC7230]  Fielding, R., Ed. and J. Reschke, Ed., "Hypertext Transfer
              Protocol (HTTP/1.1): Message Syntax and Routing",
              RFC 7230, DOI 10.17487/RFC7230, June 2014,
              <http://www.rfc-editor.org/info/rfc7230>.

13.2.  Informative References

   [HAPROXY]  Tarreau, W., "The Proxy protocol", March 2017,
              <https://www.haproxy.org/download/1.8/doc/proxy-
              protocol.txt>.

   [I-D.hoffman-dns-over-https]
              Hoffman, P. and P. McManus, "DNS Queries over HTTPS",
              draft-hoffman-dns-over-https-00 (work in progress), May
              2017.

   [I-D.ietf-dnsop-dns-wireformat-http]
              Song, L., Vixie, P., Kerr, S., and R. Wan, "DNS wire-
              format over HTTP", draft-ietf-dnsop-dns-wireformat-http-01
              (work in progress), March 2017.

   [RFC3007]  Wellington, B., "Secure Domain Name System (DNS) Dynamic
              Update", RFC 3007, DOI 10.17487/RFC3007, November 2000,
              <http://www.rfc-editor.org/info/rfc3007>.







Gillmor                 Expires November 18, 2017              [Page 18]

Internet-Draft  Demultiplexing Streamed DNS from HTTP/1.x       May 2017


   [RFC5246]  Dierks, T. and E. Rescorla, "The Transport Layer Security
              (TLS) Protocol Version 1.2", RFC 5246,
              DOI 10.17487/RFC5246, August 2008,
              <http://www.rfc-editor.org/info/rfc5246>.

   [RFC6840]  Weiler, S., Ed. and D. Blacka, Ed., "Clarifications and
              Implementation Notes for DNS Security (DNSSEC)", RFC 6840,
              DOI 10.17487/RFC6840, February 2013,
              <http://www.rfc-editor.org/info/rfc6840>.

   [RFC6895]  Eastlake 3rd, D., "Domain Name System (DNS) IANA
              Considerations", BCP 42, RFC 6895, DOI 10.17487/RFC6895,
              April 2013, <http://www.rfc-editor.org/info/rfc6895>.

   [RFC7301]  Friedl, S., Popov, A., Langley, A., and E. Stephan,
              "Transport Layer Security (TLS) Application-Layer Protocol
              Negotiation Extension", RFC 7301, DOI 10.17487/RFC7301,
              July 2014, <http://www.rfc-editor.org/info/rfc7301>.

   [RFC7540]  Belshe, M., Peon, R., and M. Thomson, Ed., "Hypertext
              Transfer Protocol Version 2 (HTTP/2)", RFC 7540,
              DOI 10.17487/RFC7540, May 2015,
              <http://www.rfc-editor.org/info/rfc7540>.

   [RFC7830]  Mayrhofer, A., "The EDNS(0) Padding Option", RFC 7830,
              DOI 10.17487/RFC7830, May 2016,
              <http://www.rfc-editor.org/info/rfc7830>.

   [RFC7858]  Hu, Z., Zhu, L., Heidemann, J., Mankin, A., Wessels, D.,
              and P. Hoffman, "Specification for DNS over Transport
              Layer Security (TLS)", RFC 7858, DOI 10.17487/RFC7858, May
              2016, <http://www.rfc-editor.org/info/rfc7858>.

Author's Address

   Daniel Kahn Gillmor
   American Civil Liberties Union
   125 Broad St.
   New York, NY  10004
   USA

   Email: dkg@fifthhorseman.net









Gillmor                 Expires November 18, 2017              [Page 19]