idnits 2.17.00 (12 Aug 2021) /tmp/idnits26781/draft-dkg-dprive-demux-dns-http-01.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** The document seems to lack a both a reference to RFC 2119 and the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords. RFC 2119 keyword, line 121: '...s that such a DNS client MAY decide to...' RFC 2119 keyword, line 299: '...e a request-line SHOULD ignore at leas...' RFC 2119 keyword, line 401: '...dating resolvers SHOULD set the CD bit...' RFC 2119 keyword, line 591: '... MAY also take one of the following ...' RFC 2119 keyword, line 614: '...types of traffic SHOULD expect a trans...' (1 more instance...) -- The draft header indicates that this document updates RFC7230, but the abstract doesn't seem to mention this, which it should. -- The draft header indicates that this document updates RFC7540, but the abstract doesn't seem to mention this, which it should. -- The draft header indicates that this document updates RFC1035, but the abstract doesn't seem to mention this, which it should. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year (Using the creation date from RFC7230, updated by this document, for RFC5378 checks: 2007-12-21) -- The document seems to lack a disclaimer for pre-RFC5378 work, but may have content which was first submitted before 10 November 2008. If you have contacted all the original authors and they are all willing to grant the BCP78 rights to the IETF Trust, then this is fine, and you can ignore this comment. If not, you may need to add the pre-RFC5378 disclaimer. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- The document date (May 03, 2017) is 1844 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- -- Looks like a reference, but probably isn't: '6' on line 453 -- Looks like a reference, but probably isn't: '7' on line 453 -- Looks like a reference, but probably isn't: '8' on line 454 -- Looks like a reference, but probably isn't: '10' on line 454 -- Looks like a reference, but probably isn't: '12' on line 454 -- Looks like a reference, but probably isn't: '9' on line 455 -- Looks like a reference, but probably isn't: '11' on line 455 -- Looks like a reference, but probably isn't: '13' on line 455 -- Looks like a reference, but probably isn't: '0' on line 606 -- Looks like a reference, but probably isn't: '1' on line 609 -- Looks like a reference, but probably isn't: '4' on line 597 -- Looks like a reference, but probably isn't: '5' on line 600 -- Looks like a reference, but probably isn't: '2' on line 603 == Outdated reference: A later version (-03) exists of draft-ietf-dnsop-dns-wireformat-http-01 -- Obsolete informational reference (is this intentional?): RFC 5246 (Obsoleted by RFC 8446) Summary: 1 error (**), 0 flaws (~~), 2 warnings (==), 19 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 dprive D. Gillmor 3 Internet-Draft ACLU 4 Updates: 1035, 7230, 7540 (if approved) May 03, 2017 5 Intended status: Informational 6 Expires: November 4, 2017 8 Demultiplexing Streamed DNS from HTTP 9 draft-dkg-dprive-demux-dns-http-01 11 Abstract 13 DNS over TCP and traditional HTTP are both stream-oriented, client- 14 speaks-first protocols. They can both be run over a stream-based 15 security protocol like TLS. A server accepting a stream-based client 16 can distinguish between a valid stream of DNS queries and valid 17 stream of HTTP requests by simple observation of the first few octets 18 sent by the client. This can be done without any external 19 demultiplexing mechanism like TCP port number or ALPN. 21 Implicit multiplexing of the two protocols over a single listening 22 port can be useful for obscuring the presence of DNS queries from a 23 network observer, which makes it relevant for DNS privacy. 25 Status of This Memo 27 This Internet-Draft is submitted in full conformance with the 28 provisions of BCP 78 and BCP 79. 30 Internet-Drafts are working documents of the Internet Engineering 31 Task Force (IETF). Note that other groups may also distribute 32 working documents as Internet-Drafts. The list of current Internet- 33 Drafts is at http://datatracker.ietf.org/drafts/current/. 35 Internet-Drafts are draft documents valid for a maximum of six months 36 and may be updated, replaced, or obsoleted by other documents at any 37 time. It is inappropriate to use Internet-Drafts as reference 38 material or to cite them other than as "work in progress." 40 This Internet-Draft will expire on November 4, 2017. 42 Copyright Notice 44 Copyright (c) 2017 IETF Trust and the persons identified as the 45 document authors. All rights reserved. 47 This document is subject to BCP 78 and the IETF Trust's Legal 48 Provisions Relating to IETF Documents 49 (http://trustee.ietf.org/license-info) in effect on the date of 50 publication of this document. Please review these documents 51 carefully, as they describe your rights and restrictions with respect 52 to this document. Code Components extracted from this document must 53 include Simplified BSD License text as described in Section 4.e of 54 the Trust Legal Provisions and are provided without warranty as 55 described in the Simplified BSD License. 57 Table of Contents 59 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 60 2. Distinguish only at the start of a stream . . . . . . . . . . 3 61 2.1. Why not ALPN? . . . . . . . . . . . . . . . . . . . . . . 4 62 3. Overview of initial octets . . . . . . . . . . . . . . . . . 4 63 3.1. DNS stream initial octets . . . . . . . . . . . . . . . . 4 64 3.2. HTTP initial octets . . . . . . . . . . . . . . . . . . . 5 65 3.2.1. HTTP/0.9 . . . . . . . . . . . . . . . . . . . . . . 6 66 3.2.2. HTTP/1.0 and HTTP/1.1 . . . . . . . . . . . . . . . . 6 67 3.2.3. HTTP/2 . . . . . . . . . . . . . . . . . . . . . . . 7 68 4. Specific octets . . . . . . . . . . . . . . . . . . . . . . . 8 69 4.1. octets 0 and 1 . . . . . . . . . . . . . . . . . . . . . 8 70 4.2. octets 2 and 3 . . . . . . . . . . . . . . . . . . . . . 8 71 4.3. octet 4 . . . . . . . . . . . . . . . . . . . . . . . . . 9 72 4.4. octet 5 . . . . . . . . . . . . . . . . . . . . . . . . . 9 73 4.5. octets 6 and 7 . . . . . . . . . . . . . . . . . . . . . 10 74 4.6. octets 8 through 11 . . . . . . . . . . . . . . . . . . . 10 75 4.7. octets 12 and 13 . . . . . . . . . . . . . . . . . . . . 10 76 5. Combinations of octets . . . . . . . . . . . . . . . . . . . 10 77 5.1. Proof: a valid DNS message cannot be an HTTP query . . . 11 78 6. Guidance for Demultiplexing Servers . . . . . . . . . . . . . 12 79 6.1. Without supporting HTTP/0.9 . . . . . . . . . . . . . . . 12 80 6.2. Supporting archaic HTTP/0.9 clients . . . . . . . . . . . 12 81 6.3. Signaling demultiplexing capacity . . . . . . . . . . . . 13 82 7. Guidance for DNS clients . . . . . . . . . . . . . . . . . . 13 83 7.1. Interpreting failure . . . . . . . . . . . . . . . . . . 14 84 8. Guidance for HTTP clients . . . . . . . . . . . . . . . . . . 15 85 9. Security Considerations . . . . . . . . . . . . . . . . . . . 15 86 10. Privacy Considerations . . . . . . . . . . . . . . . . . . . 15 87 11. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15 88 12. Document Considerations . . . . . . . . . . . . . . . . . . . 16 89 13. References . . . . . . . . . . . . . . . . . . . . . . . . . 16 90 13.1. Normative References . . . . . . . . . . . . . . . . . . 16 91 13.2. Informative References . . . . . . . . . . . . . . . . . 17 92 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 18 94 1. Introduction 96 DNS and HTTP are both client-speaks-first protocols capable of 97 running over stream-based transport like TCP, or as the payload of a 98 typical TLS [RFC5246] session. 100 There are some contexts where it is useful for a server to be able to 101 decide what protocol is used by an incoming TCP stream, to choose 102 dynamically between DNS and HTTP on the basis of the stream itself 103 (rather than a port designation or other explicit demultiplexing). 105 For example, a TLS terminator listening on port 443 might be willing 106 to serve DNS-over-TLS [RFC7858] as well as HTTPS. 108 A simple demultiplexing server should do this demuxing based on the 109 first few bytes sent by the client on a given stream; once a choice 110 has been established, the rest of the stream is committed to one or 111 the other interpretation. 113 This document provides proof that a demultiplexer can robustly 114 distinguish HTTP from DNS on the basis of the content of the stream 115 alone. 117 A DNS client that knows it is talking to a server which is this 118 position (e.g. trying to do DNS-over-TLS on TCP port 443, used 119 traditionally only for HTTPS) might also want to be aware of network 120 traffic patterns that could confuse such a server. This document 121 presents explicit mitigations that such a DNS client MAY decide to 122 use. 124 This document limits its discussion of HTTP over TCP or TLS or some 125 other classical stream-based protocol (it excludes HTTP over QUIC, 126 for example). Likewise, it considers only the TCP variant of DNS 127 (and excludes DNS over UDP or any other datagram transport). 129 FIXME: address network stack ossification here? 131 2. Distinguish only at the start of a stream 133 A server which attempts to distinguish DNS queries from HTTP requests 134 individually might consider using these guidelines in the middle of a 135 running stream (e.g. at natural boundaries, like the end of an HTTP 136 request, or after a DNS message), but this document focuses 137 specifically on a heuristic choice for the whole stream, based on the 138 initial few octets sent by the client. 140 While it's tempting to consider distinguishing at multiple points in 141 the stream, the complexities of determining the specific end of an 142 HTTP/1.1 request body, and the difficulty in distinguishing an HTTP/2 143 frame header from a streamed DNS message make this more difficult to 144 implement. Interleaving the responses themselves on a stream with 145 multiple data elements is also challenging. So do not use this 146 technique anywhere but at the beginning of a stream! 148 If being able to interleave DNS queries with HTTP requests on a 149 single stream is desired, a strategy like 150 [I-D.ietf-dnsop-dns-wireformat-http] is recommended instead. 152 2.1. Why not ALPN? 154 If this is done over TLS, a natural question is whether the client 155 should simply indicate its preferred protocol in the TLS handshake's 156 ALPN [RFC7301] extension. 158 However, ALPN headers are visible to a network observer, and a 159 network controller attempting to confine the user's DNS traffic to a 160 limited set of servers could use the ALPN header as a signal to block 161 DNS-specific streams. 163 3. Overview of initial octets 165 3.1. DNS stream initial octets 167 [RFC1035] section 4.2.2 ("TCP Usage") shows that every stream-based 168 DNS connection starts with a DNS message, preceded with a 2-octet 169 message length field: 171 The message is prefixed with a two byte length field which gives 172 the message length, excluding the two byte length field. 174 [RFC6895] section 2 represents the DNS message header section, which 175 is the first part of the DNS message on the wire (after the message 176 length). 178 1 1 1 1 1 1 179 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 180 +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ 181 | ID | 182 +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ 183 |QR| OpCode |AA|TC|RD|RA| Z|AD|CD| RCODE | 184 +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ 185 | QDCOUNT/ZOCOUNT | 186 +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ 187 | ANCOUNT/PRCOUNT | 188 +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ 189 | NSCOUNT/UPCOUNT | 190 +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ 191 | ARCOUNT | 192 +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ 194 So in a DNS over TCP stream, the interpretation of the initial 14 195 octets are fixed based on information about the first query sent on 196 the stream: 198 o 0,1: length of initial DNS message 200 o 2,3: DNS Transaction ID 202 o 4,5: DNS opcode, flags, and response code 204 o 6,7: Question count (or Zone count in UPDATE) 206 o 8,9: Answer count (or Prerequisite count in UPDATE) 208 o 10,11: Authority count (or Update count in UPDATE) 210 o 12,13: Additional RR count 212 All DNS streams sent over TCP start with at least these 14 octets. 214 3.2. HTTP initial octets 216 In an HTTP stream, the first octets sent from the client are either 217 the so-called "Simple-Request" (for HTTP/0.9), the "Request-Line" 218 (for HTTP/1.0 and HTTP/1.1), which has variable characteristics, or 219 the "connection preface" (for HTTP/2) which is a fixed string. 221 Some servers may wish to ignore the oldest of these, HTTP/0.9. 223 3.2.1. HTTP/0.9 225 [RFC1945] section 4.1 says that HTTP/0.9 queries (that is, HTTP 226 queries from before HTTP/1.0 was formalized) use this form: 228 Simple-Request = "GET" SP Request-URI CRLF 230 Note that HTTP/0.9 clients send this string and only this string, 231 nothing else (no request body, no subsequent requests). The 232 "Request-URI" token is guaranteed to start with a printable ASCII 233 character, and cannot contain any members of the CTL class (values 234 0x00 through 0x1F) but due to loose early specifications, it might 235 sometimes contain high-valued octets (those with the most-significant 236 bit set - 0x80 or above). 238 So the first 5 octets are all constrained to be no less than 0x20 239 (SP) and no more than 0x7F (DEL), and all subsequent octets sent from 240 the client have a value at least 0x0A (LF). 242 The shortest possible HTTP/0.9 client request is: 244 char: G E T SP / CR LF 245 index: 0 1 2 3 4 5 6 247 The lowest possible HTTP/0.9 client request (sorted ASCIIbetically) 248 is: 250 char: G E T SP + : CR LF 251 index: 0 1 2 3 4 5 6 7 253 3.2.2. HTTP/1.0 and HTTP/1.1 255 The request line format for HTTP/1.1 matches that of HTTP/1.0 256 (HTTP/1.1 adds protocol features like pipelining, but doesn't change 257 the request form itself). But unlike HTTP/0.9, the initial verb (the 258 "method") can vary. 260 [RFC7230] section 3.1.1 says that the first line of an HTTP/1.1 261 request is: 263 request-line = method SP request-target SP HTTP-version CRLF 264 method = token 266 and [RFC7230] section 3.2.6 says: 268 token = 1*tchar 270 tchar = "!" / "#" / "$" / "%" / "&" / "'" / "*" 271 / "+" / "-" / "." / "^" / "_" / "`" / "|" / "~" 272 / DIGIT / ALPHA 273 ; any VCHAR, except delimiters 275 and VCHAR is defined in [RFC5234] appendix B.1 as: 277 VCHAR = %x21-7E 279 "request-target" itself cannot contain 0x20 (SP) or any CTL 280 characters, or any characters above the US-ASCII range (> 0x7F). 282 And the "HTTP-version" token is either the literal string "HTTP/1.0" 283 or the literal string "HTTP/1.1", both of which are constrained to 284 the same printable-ASCII range. 286 The ASCIIbetically-lowest shortest possible HTTP/1.0 or HTTP/1.1 287 request is: 289 char: ! SP / SP H T T P / 1 . 0 CR LF CR LF 290 index: 0 1 2 3 4 5 6 7 8 9 0 a b c d e 292 In any case, no HTTP/1.0 or HTTP/1.1 request line can include any 293 values lower than 0x0A (LF) or greater than 0x7F (DEL) in the first 294 15 octets. 296 However, [RFC7230] section 3.1.1 also says: 298 In the interest of robustness, a server that is expecting to receive 299 and parse a request-line SHOULD ignore at least one empty line (CRLF) 300 received prior to the request-line. 302 So we should also consider accepting an arbitrary number of repeated 303 CRLF sequences before the request-line as a potentially-valid HTTP 304 client behavior. 306 3.2.3. HTTP/2 308 [RFC7540] section 3.5 says: 310 In HTTP/2, each endpoint is required to send a connection preface as 311 a final confirmation of the protocol in use and to establish the 312 initial settings for the HTTP/2 connection. The client and server 313 each send a different connection preface. 315 The client connection preface starts with a sequence of 24 octets, 316 which in hex notation is: 318 0x505249202a20485454502f322e300d0a0d0a534d0d0a0d0a 320 That is, the connection preface starts with the string "PRI * 321 HTTP/2.0\r\n\r\nSM\r\n\r\n"). 323 The highest valued octet here is 0x54 ("T"), and the lowest is 0x0A 324 (LF). 326 4. Specific octets 328 The sections below examine likely values of specific octet positions 329 in the stream. All octet indexes are 0-based. 331 4.1. octets 0 and 1 333 Any DNS message less than 3338 octets sent as the initial query over 334 TCP can be reliably distinguished from any version of HTTP by the 335 first two octets of the TCP stream alone. 337 3338 is 0x0D0A, or the ASCII string CRLF, which some HTTP clients 338 might send before an initial request. No HTTP client can 339 legitimately send anything lower than this. 341 Most DNS queries are easily within this range automatically. 343 4.2. octets 2 and 3 345 In a DNS stream, octets 2 and 3 represent the client-chosen message 346 ID. The message ID is used to bind messages with responses. Over 347 connectionless transports like UDP, this is an important anti- 348 spoofing measure, as well as a distinguishing measure for clients 349 reusing the same UDP port for multiple outstanding queries. Standard 350 DNS clients already explicitly randomize this value. 352 For the connection-oriented streaming DNS discussed here, the anti- 353 spoofing characteristics are not relevant (the connection itself 354 provides anti-spoofing), so the client is free to choose arbitrary 355 values. 357 With a standard DNS client which fully-randomizes these values, only 358 25% of generated queries will have the high bits of both octets set 359 to 0. 100% of all HTTP requests will have the high bits of both of 360 these octets cleared. Similarly, some small percentage of randomly- 361 generated DNS queries will have values here lower than 0x0A, while no 362 HTTP clients will ever send these low values. 364 4.3. octet 4 366 In a DNS stream, octet 4 combines several fields: 368 0 1 2 3 4 5 6 7 369 +--+--+--+--+--+--+--+--+ 370 |QR| Opcode |AA|TC|RD| 371 +--+--+--+--+--+--+--+--+ 373 In a standard DNS query sent over a streaming interface, QR, Opcode, 374 AA, and TC are all set to 0. The least-significant bit (RD - 375 Recursion Desired) is set when a packet is sent from a stub to a 376 recursive resolver. The value of such an octet is 0x01. This value 377 never occurs in octet 4 of a legitimate HTTP client. 379 But under DNS UPDATE ([RFC2136], Opcode is set to 5 and all the 380 option bits are cleared, which means this value would have 0x40 381 (ASCII '@'), which could legitimately occur in some HTTP requests at 382 this position.. 384 4.4. octet 5 386 In a DNS stream, octet 5 also combines several fields: 388 0 1 2 3 4 5 6 7 389 +--+--+--+--+--+--+--+--+ 390 |RA| Z|AD|CD| RCODE | 391 +--+--+--+--+--+--+--+--+ 393 In some DNS messages sent from a client, all these bits are 0. 394 However, section 5.7 of [RFC6840] suggests that queries may wish to 395 set the AD bit to indicate a desire to learn from a validating 396 resolver whether the resolver considers the contents to be Authentic 397 Data. 399 [RFC6840] also suggests that: 401 validating resolvers SHOULD set the CD bit on every upstream query. 403 So many queries, particularly from DNSSEC-validating DNS clients, are 404 likely to set bits 2 and 3, resulting in a value 0x30 (ASCII '0'). 405 This is usually a legitimate value for octet 5 in an HTTP request. 407 4.5. octets 6 and 7 409 In DNS, octets 6 and 7 represent the query count. Most DNS clients 410 will send one query at a time, which makes this value 0x0001. As 411 long as the number of initial queries does not exceed 0x0A0A (2570), 412 then at least one of these octets will have a value less than 0x0A. 413 No HTTP client sends an octet less than 0x0A in positions 6 or 7. 415 In DNS UPDATE, octets 6 and 7 represent the zone count. Entries in 416 the Zone section of the DNS UPDATE message are structured identically 417 to entries in the Query section of a standard DNS message. 419 4.6. octets 8 through 11 421 In streaming DNS, octets 8 through 11 represent answer counts and 422 authority counts in normal DNS queries, or Prerequisite and Update 423 counts in DNS UPDATE. Standard DNS queries will set them both 0. 424 DNS UPDATE queries are likely to include some records in these 425 sections, so they won't be all zero, but as long as no more than 2570 426 Prerequisite records and no more than 2570 Update records are sent, 427 at least one octet will have value less than 0x0A. But No HTTP 428 client sends an octet less tan 0x0A in these positions. 430 4.7. octets 12 and 13 432 In streaming DNS, octets 12 and 13 represent the number of Additional 433 RRs. When a DNS query is sent with EDNS(0), the OPT RR is accounted 434 for here. So this is often either 0x0000 or 0x0001. In a Secure DNS 435 UPDATE [RFC3007], the SIG(0) or TSIG record is also found in this 436 section, which could increase the values of these octets to 0x0002. 437 No HTTP client will send octets with these low values at these 438 positions. 440 5. Combinations of octets 442 In a DNS message, each Question in the Question section (or Zone in 443 the Zone section for DNS UPDATE) is at least 5 octets (1 octet for 444 zero-length QNAME + 2 octets for QTYPE + 2 octets for QCLASS), and 445 each RR (in the Answer, Authority, and Additional sections for normal 446 DNS queries; or in the Prerequisite, Update, and Additional sections 447 for DNS UPDATE) is at least 11 octets. And the header itself is 12 448 octets. 450 So we know that for a valid DNS stream, the first message has a size 451 of at least: 453 min_first_msg_size = 12 + 5 * (256*o[6] + o[7]) + 454 11 * (256*(o[8] + o[10] + o[12]) + 455 o[9] + o[11] + o[13]) 457 It's possible to compare this value with the expected first query 458 size: 460 first_msg_size = 256 * o[0] + o[1] 462 if "first_query_size" is less than "min_first_query_size" we can be 463 confident that the stream is not DNS. 465 5.1. Proof: a valid DNS message cannot be an HTTP query 467 For any a valid, stream-based DNS message: 469 o If there are fewer than 0x0A00 Questions then octet 6 < 0x0A. 471 o If there are fewer than 0x0A00 Answer RRs, then octet 8 < 0x0A. 473 o If there are fewer than 0x0A00 Authority RRs, then octet 10 < 474 0x0A. 476 o If there are fewer than 0x0A00 Additional RRs, then octet 12 < 477 0x0A. 479 If any of these four inequalities hold, then the packet is clearly 480 DNS, not HTTP. 482 if none of them hold, then there are at least 0x0A00 (2560) Questions 483 and 3*2560 == 7680 RRs. But: 485 12 + 5*2560 + 11*7680 == 97292 487 So the smallest possible DNS message where none of these four 488 inequalites hold is 97292 octets. But a DNS message is limited in 489 size to 65535 octets. 491 Therefore at least one of these inequalities holds, and one of the 492 first 14 octets of a DNS steam is < 0x0A. 494 But in a standard HTTP request, none of the first 14 octets can have 495 a value < 0x0A, so a valid DNS message cannot be mistaken for an HTTP 496 request. 498 6. Guidance for Demultiplexing Servers 500 Upon receiving a connection stream that might be either DNS or HTTP, 501 a server can inspect the initial octets of the stream to decide where 502 to send it. 504 6.1. Without supporting HTTP/0.9 506 A server that doesn't care about HTTP/0.9 can simply wait for the 507 first 14 octets of the client's request to come in. Then the 508 algorithm is: 510 bytestream = read_from_client(14) 511 for x in bytestream: 512 if (x < 0x0A) or (x > 0x7F): 513 return `DNS` 514 return `HTTP` 516 6.2. Supporting archaic HTTP/0.9 clients 518 A server that decides to try to support HTTP/0.9 clients has a 519 slightly more challenging task, since some of them may send fewer 520 octets than the initial DNS message, and the server shouldn't block 521 waiting for data that will never come. 523 bytestream = read_from_client(5) 524 for x in bytestream[0:5] 525 if (x < 0x0A) or (x > 0x7F): 526 return `DNS` 527 if (bytestream[0:4] != 'GET '): # not HTTP/0.9 528 bytestream += read_from_client(9) 529 for x in bytestream[5:14]: 530 if (x < 0x0A) or (x > 0x7f): 531 return `DNS` 532 return `HTTP` 533 else: # maybe HTTP/0.9 534 seen_sp = False 535 seen_high = False 536 while (len(bytestream) < 14): 537 if (seen_sp and seen_high): 538 return `DNS` 539 x = read_from_client(1) 540 bytestream += x 541 if (x > 0x7F): 542 seen_high = True 543 elif (x < 0x0A): 544 return `DNS` 545 elif (x == 0x20): 546 seen_sp = True # SP found before CRLF, not HTTP/0.9 547 elif (x == 0x0A): 548 return `HTTP` 549 return `HTTP` 551 Note that if read_from_client() ever fails to read the number of 552 requested bytes (e.g. because of EOF), then the stream is neither 553 valid HTTP nor valid DNS, and can be discarded. 555 6.3. Signaling demultiplexing capacity 557 FIXME: should there be a way for a listener to signal somehow that it 558 is willing and capable of handling both DNS and HTTP traffic? There 559 would need to be a different signaling mechanism for each stream 560 (unless the signalling is done somehow in an outer layer like TLS). 561 This is probably out-of-scope for this draft. 563 7. Guidance for DNS clients 565 Consider a DNS client that connects to a server that might be 566 interested in answering HTTP requests on the same address/port (or 567 other channel identifier). The client wants to send traffic that is 568 unambiguously DNS traffic to make it easy for the server to 569 distinguish it from inbound HTTP requests. Fortunately, this is 570 trivial to do. 572 Such a client should follow these guidelines: 574 o Send the DNS message size (a 16-bit integer) together in the same 575 packet with the full header of the first DNS message so that the 576 recipient can review as much as possible of the frame at once. 577 This is a best practice for efficient stream-based DNS anyway. 579 If the client is concerned about stream fragmentation that it cannot 580 control, and it is talking to a server that might be expecting 581 HTTP/0.9 clients, then the server might not be willing to wait for 582 the full initial 14 octets to make a decision. 584 Note that this fragmentation is not a concern for streams wrapped in 585 TLS when using modern AEAD ciphersuites. In this case, the client 586 gets to choose the size of the plaintext record, which is either 587 recovered by the server in full (unfragmented) or the connection 588 fails. 590 If the client does not have such a guarantee from the transport, it 591 MAY also take one of the following mitigating actions relating to the 592 first DNS message it sends in the stream [explanation of what the 593 server gets to see in the fragmented stream case are in square 594 brackets after each mitigation]: 596 o Ensure the first message is marked as a query (QR = 0), and it 597 uses opcode 0 ("Standard Query"). [bytestream[4] < 0x08] 599 o Ensure that the first message has RA = 0, Z = 0, and RCODE = 0. 600 [bytestream[5] == 0x00] 602 o Ensure that the high bit of the first octet of the message ID of 603 the first message is set. [bytesteam[2] > 0x7F] 605 o Send an initial short Server Status DNS message ahead of the 606 otherwise intended initial DNS message. [bytstream[0] == 0x00] 608 o Use the EDNS(0) padding option [RFC7830] to pad the first message 609 to a multiple of 256 octets. [bytestream[1] == 0x00] 611 7.1. Interpreting failure 613 FIXME: A DNS client that does not already know that a server is 614 willing to carry both types of traffic SHOULD expect a transport 615 connection failure of some sort. Can we say something specific about 616 what it should expect? 618 8. Guidance for HTTP clients 620 HTTP clients SHOULD NOT send HTTP/0.9 requests, since modern HTTP 621 servers are not required to support HTTP/0.9. Sending an HTTP/1.0 622 request (or any later version) is sufficient for a server to be able 623 to distinguish the two protocols. 625 9. Security Considerations 627 FIXME: Clients should locally validate DNSSEC (servers may still be 628 able to omit some records) 630 FIXME: if widely deployed, consider amplification for DDoS against 631 authoritative servers? 633 FIXME: consider dnssec transparency 635 FIXME: consider TLS session resumption - this counts as a new stream 636 boundary, so the multiplexing decision need not persist across 637 resumption. 639 FIXME: consider 0-RTT 641 FIXME: consider X.509 cert validation 643 FIXME: what other security considerations should clients take? 645 FIXME: what other security considerations should servers take? 647 10. Privacy Considerations 649 FIXME: DNS queries and HTTP requests can reveal potentially sensitive 650 information about the sender. 652 FIXME: consider DNS and HTTP traffic analysis - how should requests 653 or responses be padded, aggregated, or delayed given that streams are 654 multiplexed? 656 FIXME: any other privacy considerations? 658 11. IANA Considerations 660 This document does not ask IANA to make any changes to existing 661 registries. 663 However, it does update the DNS and HTTP specifications, to reflect 664 the fact that services using this demultiplexing technique may be 665 constrained in adoption of future versions of either DNS or HTTP if 666 those future versions modify either protocol in a way that breaks 667 with the distinctions documented here. 669 Future revisions of or extensions to stream-based DNS or HTTP should 670 take this demultiplexing technique into consideration. 672 12. Document Considerations 674 [ RFC Editor: please remove this section before publication ] 676 This document is currently edited as markdown. Minor editorial 677 changes can be suggested via merge requests at 678 https://gitlab.com/dkg/hddemux or by e-mail to the author. Please 679 direct all significant commentary to the public IETF DNS Privacy 680 mailing list: dns-privacy@ietf.org or to the IETF HTTP WG mailing 681 list: ietf-http-wg@w3.org 683 13. References 685 13.1. Normative References 687 [RFC1035] Mockapetris, P., "Domain names - implementation and 688 specification", STD 13, RFC 1035, DOI 10.17487/RFC1035, 689 November 1987, . 691 [RFC1945] Berners-Lee, T., Fielding, R., and H. Frystyk, "Hypertext 692 Transfer Protocol -- HTTP/1.0", RFC 1945, 693 DOI 10.17487/RFC1945, May 1996, 694 . 696 [RFC2136] Vixie, P., Ed., Thomson, S., Rekhter, Y., and J. Bound, 697 "Dynamic Updates in the Domain Name System (DNS UPDATE)", 698 RFC 2136, DOI 10.17487/RFC2136, April 1997, 699 . 701 [RFC5234] Crocker, D., Ed. and P. Overell, "Augmented BNF for Syntax 702 Specifications: ABNF", STD 68, RFC 5234, 703 DOI 10.17487/RFC5234, January 2008, 704 . 706 [RFC7230] Fielding, R., Ed. and J. Reschke, Ed., "Hypertext Transfer 707 Protocol (HTTP/1.1): Message Syntax and Routing", 708 RFC 7230, DOI 10.17487/RFC7230, June 2014, 709 . 711 [RFC7540] Belshe, M., Peon, R., and M. Thomson, Ed., "Hypertext 712 Transfer Protocol Version 2 (HTTP/2)", RFC 7540, 713 DOI 10.17487/RFC7540, May 2015, 714 . 716 13.2. Informative References 718 [I-D.ietf-dnsop-dns-wireformat-http] 719 Song, L., Vixie, P., Kerr, S., and R. Wan, "DNS wire- 720 format over HTTP", draft-ietf-dnsop-dns-wireformat-http-01 721 (work in progress), March 2017. 723 [RFC3007] Wellington, B., "Secure Domain Name System (DNS) Dynamic 724 Update", RFC 3007, DOI 10.17487/RFC3007, November 2000, 725 . 727 [RFC5246] Dierks, T. and E. Rescorla, "The Transport Layer Security 728 (TLS) Protocol Version 1.2", RFC 5246, 729 DOI 10.17487/RFC5246, August 2008, 730 . 732 [RFC6840] Weiler, S., Ed. and D. Blacka, Ed., "Clarifications and 733 Implementation Notes for DNS Security (DNSSEC)", RFC 6840, 734 DOI 10.17487/RFC6840, February 2013, 735 . 737 [RFC6895] Eastlake 3rd, D., "Domain Name System (DNS) IANA 738 Considerations", BCP 42, RFC 6895, DOI 10.17487/RFC6895, 739 April 2013, . 741 [RFC7301] Friedl, S., Popov, A., Langley, A., and E. Stephan, 742 "Transport Layer Security (TLS) Application-Layer Protocol 743 Negotiation Extension", RFC 7301, DOI 10.17487/RFC7301, 744 July 2014, . 746 [RFC7830] Mayrhofer, A., "The EDNS(0) Padding Option", RFC 7830, 747 DOI 10.17487/RFC7830, May 2016, 748 . 750 [RFC7858] Hu, Z., Zhu, L., Heidemann, J., Mankin, A., Wessels, D., 751 and P. Hoffman, "Specification for DNS over Transport 752 Layer Security (TLS)", RFC 7858, DOI 10.17487/RFC7858, May 753 2016, . 755 Author's Address 757 Daniel Kahn Gillmor 758 American Civil Liberties Union 759 125 Broad St. 760 New York, NY 10004 761 USA 763 Email: dkg@fifthhorseman.net