SPFBIS Working Group M. Kucherawy
Internet-Draft Cloudmark
Intended status: Informational April 16, 2012
Expires: October 18, 2012
Resolution of The SPF/Sender-ID Experiment
draft-ietf-spfbis-experiment-03
Abstract
In 2006 the IETF published a suite of protocol documents comprising
SPF and Sender-ID, two proposed email authentication protocols.
Because of interoperability concerns created by simultaneous use of
the two protocols by a receiver, and some concerns with Sender-ID and
compatibility with existing standards, the IESG required them to have
Experimental status and invited the community to observe their
deployments for a period of time, hoping convergence would be
possible later.
After six years, sufficient experience and evidence have been
collected that the experiment thus created can be considered
concluded, and a single protocol can be advanced. This memo presents
those findings.
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 October 18, 2012.
Copyright Notice
Copyright (c) 2012 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
Kucherawy Expires October 18, 2012 [Page 1]
Internet-Draft SPF/Sender-ID Experiment April 2012
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
2. Evidence of Deployment . . . . . . . . . . . . . . . . . . . . 3
2.1. DNS Resource Record Types . . . . . . . . . . . . . . . . 3
2.2. Implementations . . . . . . . . . . . . . . . . . . . . . 4
2.3. The SUBMITTER SMTP Extension . . . . . . . . . . . . . . . 5
3. Evidence of Differences . . . . . . . . . . . . . . . . . . . 5
4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.1. From the Evidence . . . . . . . . . . . . . . . . . . . . 6
4.2. Recommendations to the IESG . . . . . . . . . . . . . . . 7
5. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 7
6. Security Considerations . . . . . . . . . . . . . . . . . . . 7
7. Informative References . . . . . . . . . . . . . . . . . . . . 7
Appendix A. Experiences Developing SPF . . . . . . . . . . . . . 8
Appendix B. Acknowledgments . . . . . . . . . . . . . . . . . . . 9
Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 10
Kucherawy Expires October 18, 2012 [Page 2]
Internet-Draft SPF/Sender-ID Experiment April 2012
1. Introduction
In April 2006, the IETF published the [SPF] and Sender-ID email
authentication protocols, the latter consisting of three documents
([SUBMITTER], [SENDER-ID], and [PRA]). Both of these enable one to
publish via the Domain Name System a policy declaring which mail
servers were authorized to send email on behalf of a specific domain
name. The two protocols made use of this policy statement and some
specific (but different) logic to evaluate whether or not the email
client sending or relaying a message was authorized to do so.
Because Sender-ID supported use of the same policy statement as SPF,
the IESG at the time was concerned that an implementation of
Sender-ID might erroneously apply that statement to a message and,
depending on selected recipient actions, could improperly interfere
with message delivery. As a result, the IESG required the
publication of all of these documents as Experimental, and requested
that the community observe deployment and operation of the protocols
over a period of two years from publication in order to determine a
reasonable path forward. (For further details about the IESG's
concern, see the IESG Note prepended to all of those documents.)
Accordingly, this memo resolves the experiment by presenting evidence
regarding both deployment and efficacy of the two protocols. At the
end it presents conclusions based on the data collected.
It is important to note that this memo makes no direct technical
comparison of the two protocols in terms of correctness, weaknesses,
or use case coverage. The email community at large has already done
that. Rather, the analysis presented here merely observes what has
been deployed and supported in the time since the protocols were
published, and draws conclusions based on those observations.
2. Evidence of Deployment
This section presents the collected research done to determine what
parts of the two protocol suites are in general use, as well as
related issues like DNS support.
2.1. DNS Resource Record Types
Two large-scale DNS surveys were run that looked for the two
supported kinds of resource records (RR) that can contain SPF policy
statements.
One data source for this report requested SPF records from one
million domains. The results showed 397,535 domains included non-
Kucherawy Expires October 18, 2012 [Page 3]
Internet-Draft SPF/Sender-ID Experiment April 2012
error and non-empty replies. Of these, 397,511 included type 16 (DNS
RR TXT) replies, 6,627 included type 99 (DNS RR SPF) replies, and
6,603 included both types. Of those answers retrieved, 5,291
included records that start with the string "spf2.0/" which are
specific requests for Sender-ID processing by receivers.
Another source requested SPF records from 256,295 domains for which
there was a history of previous completed SPF evaluations. Of these,
139,748 returned type 16 answers, 2,679 returned type 99 answers,
2,507 returned both types, and the rest returned neither. Of those
answers retrieved, 6,887 included records that start with the string
"spf2.0/" which are specific requests for Sender-ID processing by
receivers.
During this second survey, some domains were observed to provide
immediate answers for type 16 queries, but would time out waiting for
replies to type 99 queries. For example, it was observed that 3,953
distinct domains in the survey returned a result of some kind (a
record or an error) for the TXT query in time N, while the SPF query
ultimately failed but only after at least time 4N.
A third source reported that of approximately 100,000 domains
surveyed, 47.7% advertised some kind of policy using SPF or
Sender-ID. Of those that did, 96.9% advertised type 16 records only,
0.2% advertised type 99 records only, and 2.9% advertised both.
A survey was done of queries for type 16 and type 99 records by
observing nameserver logs. Only a few queries were ever received for
type 99 records, and those almost exclusively came from one large
email service provider that queried for both types. The vast
majority of other querying agents only ever requested type 16.
2.2. Implementations
It is likely impossible to determine from a survey which MTAs (Mail
Transfer Agents) have SPF and/or Sender-ID checking enabled at
message ingress since it does not appear, for example, in the reply
to the EHLO command from extended [SMTP]. We therefore rely on
evidence found via web searches, and observed the following:
o A web site [SID-IMPL] dedicated to highlighting Sender-ID
implementations last updated in late 2007 listed 13
implementations, which we assume means they implement the PRA
checks. At least one of them is known no longer to be supported
by its vendor.
o The [OPENSPF] web site maintains a list of known implementations
of SPF. At the time of this memo's writing it listed six
Kucherawy Expires October 18, 2012 [Page 4]
Internet-Draft SPF/Sender-ID Experiment April 2012
libraries, 22 MTAs with built-in SPF implementations, and numerous
patches for MTAs and mail clients.
2.3. The SUBMITTER SMTP Extension
In a review of numerous MTAs in current or recent use, only two
(Santronics WinServer and McAfee MxLogic) were found to contain
implementations of the SMTP SUBMITTER extension as part of the MTA
service, which could act as an enabler to Sender-ID.
An unknown number of clients implement SUBMITTER. Although there is
substantial activity showing its use in MTA logs, it is not possible
to determine whether they are multiple instances of the same client,
or separate client implementations.
An active survey was done of a large number of running and publicly
reachable MTAs. Fewer than 4.5% of these advertised the SUBMITTER
extension. Based on the SMTP banner presented upon connection, the
entire set of SUBMITTER-enabled MTAs consisted of the two found
during the review (above) and a third whose identity could not be
positively determined.
The vast majority of the MTAs advertising SUBMITTER were different
instances of one MTA. That MTA was a mail filtering service, which
reports that about 11% of all observed SMTP sessions involve clients
that make use of SUBMITTER.
3. Evidence of Differences
It is plain from inspection of the two protocols that they have much
in common: For a single message, both require the same number of DNS
queries, and both require the same code to parse the result. The PRA
algorithm applied by Sender-ID is, however, more expensive than
simply extracting the domain name from the omnipresent
RFC5321.MailFrom. Thus, SPF is cheaper to apply to a message.
Another notable difference between the two is that the PRA part of
Sender-ID cannot be determined without proceeding to the end of the
DATA phase of an SMTP session, while SPF only requires the HELO/EHLO
and MAIL FROM command parameters. Because MTAs historically consume
substantially more resources for a transaction once past the MAIL
FROM command in the SMTP sequence, this means SPF has a substantially
lower cost to evaluate.
Separate surveys compared the cases where the PRA (used by Sender-ID)
and the RFC5321.MailFrom address (used by SPF) differed. The results
of these tests showed that at least 50% of the time the two addresses
Kucherawy Expires October 18, 2012 [Page 5]
Internet-Draft SPF/Sender-ID Experiment April 2012
were the same, but beyond that the percentage varied substantially
from one sampling location to the next due to the nature of the mail
streams they each receive.
Despite this, one set of specific data collected by a working group
contributor shows that in more than 95% of cases, Sender-ID and SPF
reach the same conclusion about a message, meaning either both
protocols return a "pass" result or both return a "fail" result. The
data set yielding this response could not further characterize the
cases in which the answers differed.
4. Conclusions
It is standard procedure within the IETF to document as standard
those protocols and practices that have come into sufficient common
use as to become part of the basic infrastructure.
4.1. From the Evidence
Given the six years that have passed since the publication of the
experimental RFCs, and the evidence reported in the earlier sections
of this document, the following conclusions are supported:
1. There has not been substantial adoption of the type 99 (SPF) DNS
resource record. In all large-scale surveys performed for this
work, less than 2% of responding domains published type 99
records, and almost no clients requested them.
2. Of the records retrieved, fewer than 5% requested processing of
messages using the PRA algorithm, which was an integral part of
Sender-ID.
3. Although the two mechanisms often used different email addresses
as the subject being evaluated, no data collected showed any
substantial operational benefit (e.g., cheaper processing,
improved accuracy) to using Sender-ID over SPF.
4. A review of known implementations shows significant support for
both protocols, though there were more implementations in support
of SPF than of Sender-ID. Further, the SPF implementations
showed better upkeep and current interest than the Sender-ID
implemenations.
5. A survey of running MTAs shows fewer than 5% of them advertised
the SUBMITTER extension, which is a Sender-ID enabler. Only
three implementations of it were found.
Kucherawy Expires October 18, 2012 [Page 6]
Internet-Draft SPF/Sender-ID Experiment April 2012
6. The overall lack of Sender-ID adoption likely has as a
contributing factor the additional cost of having to accept and
process the entire message rather than just a portion of its
envelope.
7. Although they may be marginal, there remain obstacles to
deployment of protocols that use DNS RRtypes other than the most
common ones, including firewalls and DNS servers that block or
discard requests for unknown types. Further, few if any web-
based DNS configuration tools offer support for type 99 records.
4.2. Recommendations to the IESG
In light of the above, the working group recommends to the IESG the
following:
1. that the experiment comprising the series of RFCs defining the
SUBMITTER SMTP extension, the Sender-ID mechanism, the Purported
Responsible address algorithm, and SPF, be considered concluded;
2. that continued widespread use of [SPF] indicates it is worthy of
consideration for the Standards Track;
3. that the absence of significant adoption of the type 99 DNS
resource record, the [SUBMITTER] extension, [SENDER-ID], and
[PRA], indicates that they are de facto obsolete.
Appendix A is offered as a cautionary review of problems that
affected the process of developing SPF and Sender-ID in terms of
their use of the DNS.
5. IANA Considerations
This memo presents no actions for IANA. [RFC Editor: Please remove
this section prior to publication.]
6. Security Considerations
This memo contains information for the community only, akin to an
implementation report, and does not introduce any new security
concerns. Its implications could, in fact, resolve some.
7. Informative References
[OPENSPF] "Sender Policy Framework: Project Overview",
Kucherawy Expires October 18, 2012 [Page 7]
Internet-Draft SPF/Sender-ID Experiment April 2012
<http://www.openspf.net>.
[PRA] Lyon, J., "Purported Responsible Address in E-Mail
Messages", RFC 4407, April 2006.
[SENDER-ID]
Lyon, J. and M. Wong, "Sender ID: Authenticating E-Mail",
RFC 4406, April 2006.
[SID-IMPL]
"Sender ID Framework Industry Support and Solutions",
October 2007, <http://www.microsoft.com/mscorp/safety/
technologies/senderid/support.mspx>.
[SMTP] Klensin, J., "Simple Mail Transfer Protocol", RFC 5321,
October 2008.
[SPF] Wong, M. and W. Schlitt, "Sender Policy Framework (SPF)
for Authorizing Use of Domains in E-Mail, Version 1",
RFC 4408, April 2006.
[SUBMITTER]
Allman, E. and H. Katz, "SMTP Service Extension for
Indicating the Responsible Submitter of an E-Mail
Message", RFC 4405, April 2006.
Appendix A. Experiences Developing SPF
SPF was originally developed by a community of interested developers
outside the IETF. It was brought to the IETF for standardization
only after the specification was relatively mature and ready for the
rigors of the IETF publication process.
At the time of SPF's initial development, the prospect of getting a
DNS resource record (RR) type allocated for SPF was not seriously
considered, partly because it was perceived to have high barriers to
entry. As a result, by the time the working group was formed, there
was already a substantial and growing installed base that had SPF
running using TXT RRs. Eventually the application was made for the
new RR type as a result of pressure from the DNS experts in the
community, who encouraged doing so as the preferred path toward using
the DNS for storing such things as policy data.
Later, after type 99 was assigned (long after IESG approval of the
document, in fact), a plan was put into place to effect a gradual
transition to using type 99 instead of using type 16. This plan
failed to take effect for four primary reasons:
Kucherawy Expires October 18, 2012 [Page 8]
Internet-Draft SPF/Sender-ID Experiment April 2012
1. there was hesitation to make the transition because of concerns
that nameservers (and, in fact, DNS-aware firewalls) would drop
or reject requests for unknown RR types (see Section 2 for
evidence of this), which means successful rollout of a new RR
type is contingent upon widespread adoption of updated
nameservers and resolver functions;
2. many DNS provisioning tools (e.g., web interfaces to controlling
DNS zone data) were, and still are, typically lethargic about
adding support for new RR types;
3. the substantial deployed base was already using type 16, and it
was working just fine, leading to inertia;
4. [SPF] itself included a faulty transition plan: It said a server
SHOULD publish both types and MUST publish at least one, while a
client can query either or both, which means both can claim to be
fully compliant while failing utterly to interoperate.
It is likely that this will happen again if the bar to creating new
RR types even for experimental development purposes is not lowered,
and handling of unknown RR types in software becomes generally more
graceful. Also important in this regard is encouragement of support
for new RR types in DNS record provisioning tools.
There are DNS experts within the community that will undoubtedly
point to DNS servers and firewalls that mistreat queries for unknown
RR types, and claim they are broken, as a way of answering this
concern. This is undoubtedly correct, but the reality is that they
are among us and likely will be for some time, and this needs to be
considered as new protocols and IETF procedures are developed.
Appendix B. Acknowledgments
The following provided operational data that contributed to the
evidence presented above:
Cisco: contributed data about observed Sender-ID and SPF records in
the DNS for a large number of domains
Hotmail: contributed data about the difference between
RFC5321.MailFrom and RFC5322.From domains across large mail
volumes, and a survey of DNS queries observed in response to
outgoing mail traffic
Kucherawy Expires October 18, 2012 [Page 9]
Internet-Draft SPF/Sender-ID Experiment April 2012
John Levine: conducted a survey of DNS server logs to evaluate SPF-
related query traffic
McAfee: provided details about their SUBMITTER implementation and
usage statistics
Santronics: contributed data about the use of the SUBMITTER
extension in aggregate SMTP client traffic
The Trusted Domain Project: contributed data about the difference
between Sender-ID and SPF results, conducted one of the two
detailed TXT/SPF record surveys including collecting timing data,
and conducted the MTA SUBMITTER survey
The author would also like to thank the following for their
contributions to the development of the text in this memo: Dave
Crocker, Scott Kitterman, Barry Leiba, John Levine, Hector Santos,
and Alessandro Vesely.
Author's Address
Murray S. Kucherawy
Cloudmark
128 King St., 2nd Floor
San Francisco, CA 94107
USA
Phone: +1 415 946 3800
Email: msk@cloudmark.com
Kucherawy Expires October 18, 2012 [Page 10]