idnits 2.17.1 draft-ietf-dkim-implementation-report-04.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 : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (November 7, 2010) is 4919 days in the past. Is this intentional? -- Found something which looks like a code comment -- if you have code sections in the document, please surround them with '' and '' lines. Checking references for intended status: Informational ---------------------------------------------------------------------------- ** Obsolete normative reference: RFC 4871 (ref. 'DKIM') (Obsoleted by RFC 6376) Summary: 1 error (**), 0 flaws (~~), 1 warning (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 DKIM Working Group M. Kucherawy 3 Internet-Draft Cloudmark 4 Intended status: Informational November 7, 2010 5 Expires: May 11, 2011 7 RFC4871 Implementation Report 8 draft-ietf-dkim-implementation-report-04 10 Abstract 12 This document contains an implementation report for the IESG covering 13 DomainKeys Identified Mail (DKIM) in support of the advancement of 14 that specification along the Standards Track. 16 Status of this Memo 18 This Internet-Draft is submitted in full conformance with the 19 provisions of BCP 78 and BCP 79. 21 Internet-Drafts are working documents of the Internet Engineering 22 Task Force (IETF). Note that other groups may also distribute 23 working documents as Internet-Drafts. The list of current Internet- 24 Drafts is at http://datatracker.ietf.org/drafts/current/. 26 Internet-Drafts are draft documents valid for a maximum of six months 27 and may be updated, replaced, or obsoleted by other documents at any 28 time. It is inappropriate to use Internet-Drafts as reference 29 material or to cite them other than as "work in progress." 31 This Internet-Draft will expire on May 11, 2011. 33 Copyright Notice 35 Copyright (c) 2010 IETF Trust and the persons identified as the 36 document authors. All rights reserved. 38 This document is subject to BCP 78 and the IETF Trust's Legal 39 Provisions Relating to IETF Documents 40 (http://trustee.ietf.org/license-info) in effect on the date of 41 publication of this document. Please review these documents 42 carefully, as they describe your rights and restrictions with respect 43 to this document. Code Components extracted from this document must 44 include Simplified BSD License text as described in Section 4.e of 45 the Trust Legal Provisions and are provided without warranty as 46 described in the Simplified BSD License. 48 Table of Contents 50 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 51 2. Definitions . . . . . . . . . . . . . . . . . . . . . . . . . 4 52 3. DKIM Interoperability Event . . . . . . . . . . . . . . . . . 5 53 3.1. Participants . . . . . . . . . . . . . . . . . . . . . . . 5 54 3.2. Testing Methodology . . . . . . . . . . . . . . . . . . . 5 55 3.3. Observations . . . . . . . . . . . . . . . . . . . . . . . 5 56 3.4. Results . . . . . . . . . . . . . . . . . . . . . . . . . 6 57 4. Collected DKIM Interoperability and Use Data . . . . . . . . . 8 58 4.1. The OpenDKIM Project . . . . . . . . . . . . . . . . . . . 8 59 4.1.1. Details . . . . . . . . . . . . . . . . . . . . . . . 8 60 4.1.2. Results . . . . . . . . . . . . . . . . . . . . . . . 8 61 4.1.3. Conclusions . . . . . . . . . . . . . . . . . . . . . 10 62 4.2. AOL, Inc. Data . . . . . . . . . . . . . . . . . . . . . . 11 63 4.3. Google Mail Data . . . . . . . . . . . . . . . . . . . . . 11 64 5. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 13 65 6. Security Considerations . . . . . . . . . . . . . . . . . . . 14 66 7. References . . . . . . . . . . . . . . . . . . . . . . . . . . 15 67 7.1. Normative References . . . . . . . . . . . . . . . . . . . 15 68 7.2. Informative References . . . . . . . . . . . . . . . . . . 15 69 Appendix A. Acknowledgements . . . . . . . . . . . . . . . . . . 16 70 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 17 72 1. Introduction 74 DomainKeys Identified Mail (DKIM), published in May 2007, has reached 75 a level of maturity sufficient to consider its advancement along the 76 standards track. Enclosed is a summary of collected interoperability 77 data provided from sources that are aggregating such information as 78 well as from a more formal DKIM interoperability event that took 79 place in October 2007. 81 2. Definitions 83 DomainKeys Identified Mail is defined in [DKIM]. 85 Various terms specific to email are used in this document. Their 86 definitions and further discussion can be found in [EMAIL-ARCH]. 88 3. DKIM Interoperability Event 90 In October 2007, Alt-N Technologies of Dallas, Texas hosted an 91 interoperability and testing event at their headquarters. Twenty 92 organizations sent engineers and their various DKIM implementations 93 to connect to a private internal network and exchange test messages 94 and tabulate observed results. 96 3.1. Participants 98 The interoperability event included participants from all of the 99 following organizations: Alt-N Technologies, AOL Inc., AT&T 100 Laboratories, Bizanga Ltd., Brandenburg InternetWorking, Brandmail 101 Solutions, ColdSpark, Constant Contact, Inc., DKIMproxy, Domain 102 Assurance Council, Google Inc., ICONIX Inc., Internet Initiative 103 Japan (IIJ), Ironport Systems, Message Systems, Port25 Solutions, 104 Postfix, Sendmail, Inc., StrongMail Systems, and Yahoo! Inc. Most of 105 the participants traveled to Dallas and participated in person, but a 106 few operated remotely. 108 Nearly all of the implementations were based on disjoint code 109 development projects. A few were based on a common open source base 110 project. 112 3.2. Testing Methodology 114 Participants were encouraged before the event to craft a set of test 115 messages meant to exercise their own implementations as well as those 116 of the other participants, both in terms of successful verifications 117 as well as some expected to fail. Test cases were developed with the 118 intent of confounding verifiers that may not have implemented the 119 [ABNF] of [DKIM] correctly. 121 The participants set up Mail Transfer Agents (MTAs) equipped with 122 their own DKIM signing and verifying modules, and their own tools to 123 generate mail to be signed along with tools to analyze the results 124 post-verification. They then sent their own batteries of test 125 messages, looking for both expected and unexpected failures in 126 response. Some implementations included "auto-responders" that would 127 reply with verification results, while others simply collected the 128 results that would then be shared manually. 130 3.3. Observations 132 All of the packages implemented all of the required portions of 133 [DKIM] in terms of both signature and key features. Most of the 134 packages implemented all of the optional features of both signatures 135 and keys. There were at least two implementations of each optional 136 feature. 138 The interoperability testing was largely successful. As might be 139 expected, there were many verification false negatives or false 140 positives that were the result of bugs in corner cases of some of the 141 implementations presented for testing. In such cases the developers 142 were able to identify the issue as resulting from their own mis- 143 reading of the specification and not an error in the specification 144 itself. 146 Several of the failures did occur as a result of specification 147 ambiguities. The participants discussed each of these in turn and 148 were able to come to consensus on how they believed the specification 149 should be changed to resolve them. 151 The participants agreed to keep the results about the specific tests 152 private. Accordingly, those data are not presented here. 154 3.4. Results 156 The handful of interoperability issues described above that referred 157 to weaknesses or ambiguities in [DKIM] resulted in several errata 158 being opened via the RFC Editor web site. These are being addressed 159 in an RFC4871bis draft effort that is now starting from within the 160 DKIM working group. 162 The errata items, in summary: 164 o explicit canonicalized forms of empty bodies for each 165 canonicalization method, along with their SHA1 and SHA256 hash 166 values (errata #1376 and #1377) 168 o clarification about normative text regarding the "a=" tag (errata 169 #1378) 171 o ABNF corrections regarding the "z=" tag (errata #1379) 173 o informative discussion regarding the "x=" tag (errata #1380) 175 o normative clarifications about "q=", "h=", "k=", "s=" and "t=" 176 tags (errata #1381 and #1382) 178 o correction of "g=" description to match its ABNF (errata #1383) 180 o clarifications about "relaxed" body canonicalization (errata 181 #1384) 183 o correction to the signature example (errata #1386) 185 o ABNF corrections regarding the "h=" tag (errata #1461) 187 o ABNF corrections regarding the "v=" tag (errata #1487) 189 o discussion of DomainKeys compatibility (errata #1532) 191 o discussion about what constitutes the actual value of the "b=" tag 192 (errata #1596) 194 4. Collected DKIM Interoperability and Use Data 196 Several implementations are collecting private data about DKIM use, 197 signature survivability, which properties of the base specification 198 are observed in public use, etc. This section includes collection 199 methods and summary reports provided by those implementations. 201 4.1. The OpenDKIM Project 203 The OpenDKIM Project (http://www.opendkim.org) is an open source 204 project providing a DKIM support library, an email filter for use 205 with Mail Transfer Agents (MTAs), and a set of tools to assist with 206 deployment of DKIM. 208 4.1.1. Details 210 Recent releases have included an optional feature to record 211 statistics about messages with and without DKIM signatures. Sites 212 enabling this feature can choose to share the data with the project's 213 development team as part of this interoperability report work. The 214 data can be anonymized to conceal the sending domain and client IP 215 addresses, though these data are passed through a one-way hash to 216 enable collation of data from common sources. 218 4.1.2. Results 220 At the time of writing of this document, five weeks of data had been 221 collected. The results of this effort are as follows: 223 Reporting Hosts: six individual MTAs representing four distinct 224 ADMDs 226 Total Messages: 2558218 228 Signatures: 1869088 messages (73.0%) were not signed; 676133 (26.4%) 229 had one signature; 12906 (0.5%) had two signatures; the remainder 230 (less than 0.01%) had more. 232 Signing Algorithms: 50.5% of signatures used "rsa-sha1", while the 233 balance used "rsa-sha256". 235 Header Canonicalization Algorithms: 14.7% of signatures used 236 "simple", while the balance used "relaxed"; when grouped by 237 domains, the percentages were similar. 239 Body Canonicalization Algorithms: 26.9% of signatures used "simple", 240 while the balance used "relaxed"; 18.9% of observed signing 241 domains used "simple" while the balance used "relaxed". 243 Keys in Test Mode: 46.6% of keys retrieved from the DNS were tagged 244 as being in test mode. 246 Keys with Specific Granularity: 14 keys were retrieved that had 247 specific names in their "g=" tags. 249 Keys with Syntax Errors: Less than 0.1% of keys retrieved from the 250 DNS had syntax errors. 252 DomainKeys Compatibility: 1.2% of the retrieved keys appeared to be 253 intended for use with the older DomainKeys proposal rather than 254 DKIM 256 Missing Keys: 1.7% of signatures received referenced keys that were 257 not found in the DNS 259 Optional Signature Tags: Of the optional signature tags supported by 260 the base specification, "t=" was seen 46.6% of the time (1% of 261 which included timestamps in the future, even after forgiving some 262 clock drift); "x=" was seen 4.2% of the time; "l=" was seen 4% of 263 the time; "z=" was seen 7.2% of the time. 265 Body Length Limits: Of the signatures for which "l=" was used, 8.4% 266 of them signed none of the body, and 84.6% of the rest had the 267 body extended after signing. 269 Signature Pass Rates: Overall, 92% of observed signatures were 270 successfully verified. 272 Pass Rates for Non-List Mail: Where "list mail" is defined as any 273 mail bearing one of the header fields defined in [LIST-ID] or in 274 [LIST-URLS], or a "Precedence: list" field, selecting only for 275 mail that is not list mail revealed a successful verification rate 276 of 94.9%; selecting only for list mail produced a 87.8% success 277 rate. 279 DNSSEC: No signed keys were reported in the accumulated data to 280 date. 282 Common errors: The top five verification errors observed: Key not 283 found in DNS (21.2%), key granularity mismatch (16%), DNS 284 retrieval failure such as timeouts (2.1%), key type unknown 285 (2.0%), key syntax error (1.0%). 287 Detected Header Field Changes: A subset of the reporting sites are 288 capable of reporting header fields known to have been changed in 289 transit (e.g. when "z=" tags were used by the signer). In such 290 cases, changes to the "To:" field were more common than any other 291 by almost an order of magnitude. 293 Most Commonly Signed Fields: From: (100%), To: (95.4%), Subject: 294 (95.2%), Date: (94.6%), MIME-Version: (91.3%), Content-Type: 295 (82.9%), Message-Id: (75.6%), Received: (51.8%). All others are 296 below 50%. 298 Identities: 74.7% of the signatures observed included a "d=" value 299 matching the domain in the From: field. 301 Multiple-use Signing Domains: 24789 unique signing domains were 302 observed. Of these, 32.9% of them sent a single signed message in 303 the sample period, 16.6% sent two and 9.2% sent three. 305 4.1.3. Conclusions 307 The results of the OpenDKIM work are updated constantly as more data 308 feeds come online and more data are reported. Based on the data 309 available at the time of writing, some conclusions are possible. 311 At least some implementations of all of the optional signature 312 features, all of the canonicalization combinations and all of the 313 signing algorithms are in general use. None of the features had zero 314 use counts. 316 Overall signature pass rates are generally quite high. The impact of 317 signature survivability when correlated against Mailing List Manager 318 (MLM) activity is detectable based on observed data. More research 319 into this is recommended. The DKIM Working Group is already working 320 on an Informational draft to discuss those issues. 322 That the "To" field is the one most often associated with 323 verification failures suggests some MTAs handling the message are 324 correcting cases where the field is improperly formed. A common case 325 is failing to quote the comment portion of that field when required 326 to do so by [MAIL]. Such corrections cause signatures to become 327 invalid. 329 The counts of low-use signing domains suggest that spammers, who 330 typically rotate domain names with high frequency, have adopted DKIM 331 as a tool to try to get through message filters. 333 4.2. AOL, Inc. Data 335 A one-day summary of observed traffic from AOL, Inc. reports the 336 following: 338 Ratio of DKIM-signed mail: 42% 340 Properly formed signatures: 1.4 billion 342 Malformed signatures: 3000 344 Unique signing domains: 50,000-90,000 346 Key retrieval errors: 14 million (1%) 348 Signature refers to nonexistent domain: 10 million (0.7%) 350 Signature refers to nonexistent key: 36 million (2.5%) 352 Signature refers to revoked key: 138,000 (~0% ) 354 Verified signatures: 1.2 billion (85.7%) 356 AUID matches From: domain: 1.2 billion (85.7%) 358 Failed signatures (body changed): 78 million (5.6%) 360 Failed signatures (other): 34 million (2.4%) 362 Expired signatures: less than 1 million (~0%) 364 4.3. Google Mail Data 366 Google Mail reports the following: 368 Unsigned mail: 72.1% 370 AUID matches From: domain: 68.7% 372 Signed mail that verified: 14.7% 374 Signed mail that verified in test mode: 11.7% 376 Signed mail that failed: 0.2% 377 Signed mail that failed in test mode: 0.2% 379 Body hash mismatch: 0.5% 381 Signature missing required parameters: 0.3% 383 Granularity mismatch: 0.2% 385 These data are reported based on an implementation that only 386 evaluates one signature per message. 388 All other reportable anomalies occurred in vanishingly small 389 percentages. 391 5. IANA Considerations 393 This memo contains no actions for IANA. 395 6. Security Considerations 397 This document is an implementation report and thus has no security 398 considerations. 400 7. References 402 7.1. Normative References 404 [DKIM] Allman, E., Callas, J., Delany, M., Libbey, M., Fenton, 405 J., and M. Thomas, "DomainKeys Identified Mail (DKIM) 406 Signatures", RFC 4871, May 2007. 408 [EMAIL-ARCH] 409 Crocker, D., "Internet Mail Architecture", RFC 5598, 410 July 2009. 412 7.2. Informative References 414 [ABNF] Crocker, D. and P. Overell, "Augmented BNF for Syntax 415 Specifications: ABNF", RFC 5234, January 2008. 417 [LIST-ID] Chandhok, R. and G. Wenger, "List-Id: A Structured Field 418 and Namespace for the Identification of Mailing Lists", 419 RFC 2919, March 2001. 421 [LIST-URLS] 422 Neufeld, G. and J. Baer, "The Use of URLs as Meta-Syntax 423 for Core Mail List Commands and their Transport through 424 Message Header Fields", RFC 2369, July 1998. 426 [MAIL] Resnick, P., "Internet Message Format", RFC 5322, 427 October 2008. 429 Appendix A. Acknowledgements 431 The author wishes to acknowledge the following for their review and 432 constructive criticism of this document: Dave Crocker, Tony Hansen, 433 Jeff Macdonald, S. Moonesamy and Rolf Sonneveld. 435 The author also wishes to acknowledge Margot Koschier of AOL, Inc., 436 Monica Chew of Google, and the members of The OpenDKIM Project for 437 providing data used in this report. 439 The working group expresses its profound thanks to Alt-N Technologies 440 for graciously hosting the 2007 DKIM interoperability event. 442 Author's Address 444 Murray S. Kucherawy 445 Cloudmark 446 128 King St., 2nd Floor 447 San Francisco, CA 94107 448 US 450 Phone: +1 415 946 3800 451 Email: msk@cloudmark.com