Sender ID: Difference between revisions

→External links: replaced old "cat fight" by newer IAB appeal
→See also: added "Principles of Operation"
Line 4: Line 4:


The [[MARID]] [[IETF]] working group stopped their work on the proposed standard on [[September 23]] [[2004]] because of various problems and major differences between the members of the group.
The [[MARID]] [[IETF]] working group stopped their work on the proposed standard on [[September 23]] [[2004]] because of various problems and major differences between the members of the group.
== Principles of Operation ==
Syntactically SPF and SenderID are almost identical:
Replacing the string <tt>v=spf1</tt> in a valid SPF
poiicy by one of...
# <tt>spf2.0/mfrom</tt>
# <tt>spf2.0/mfrom,pra</tt>
# <tt>spf2.0/pra,mfrom</tt>
# <tt>spf2.0/pra</tt>
...yields a syntactically valid SenderID policy, and
vice versa. The evaluation of SPF and SenderID policies
follows the same general rules. The core SenderID
specification has a normative reference to the SPF
specification and thereby inherits exactly the same error
handling, the same processing limits, and the same syntax
details as SPF.
Semantically there are only two differences: SenderID offers
the feature of ''positional'' modifiers not supported in SPF.
In practice so far no ''positional'' modifier was specified,
let alone supported in any SenderID implementation.
Except from this detail <tt>spf2.0/mfrom</tt> is semantically
the same as SPF, ''mfrom'' stands for MAIL FROM, the checked
identity in SPF. Because SPF is wider deployed than the later
proposed <tt>spf2.0/mfrom</tt> publishers of sender policies
wishing to protect their address in MAIL FROM Return-Paths
generally prefer the classic <tt>v=spf1</tt> format.
Both <tt>spf2.0/mfrom,pra</tt> and <tt>spf2.0/pra,mfrom</tt>
have the same meaning, they introduce a policy that can be
used to check the ''mfrom'' as well as the ''pra'' identity.
Last but not least <tt>spf2.0/pra</tt> introduces policies
checked only with the ''pra'' identity.  At this time it is
not specified how SPF's <nowiki>%{h}</nowiki> macro for the
HELO identity should be handled in ''pra'' checks, SenderID
implementations could handle it as syntax error.
The Purported Responsible Address or ''pra'' is determined
by a tricky evaluation of the four mail header fields...
* <tt>From</tt>,
* <tt>Sender</tt>,
* <tt>Resent-From</tt>,
* <tt>Resent-Sender</tt>.
...finally resulting either in one address, the ''pra'', or
an error for illegal combinations as explained in [[RFC]]
[http://tools.ietf.org/html/2822 2822].
Simplified the rules are <tt>Sender</tt> beats <tt>From</tt>
and <tt>Resent-</tt>* beats <tt>Sender</tt>. Note that it's
illegal to have more than one <tt>From</tt>-address without a
single <tt>Sender</tt>-address.
This ''pra'' scheme has the advantage to deal with addresses
in header fields normally displayed by MUAs (mail user agents),
unlike SPF's Return-Path (''mfrom'' in SenderID terminology).
The ''pra'' tries to counter the problem of ''phishing'',
while SPF or ''mfrom'' tries to counter the problem of spam
bounces and other auto-replies to forged Return-Paths. Two
different problems with two different proposed solutions.
OTOH ''pra'' has the severe disadvantage that forwarders and
mailing lists can only support it by modifying the mail header,
e.g. insert a <tt>Sender</tt> or <tt>Resent-Sender</tt>. The
latter violates RFC 2821 and can be incompatible with RFC 821.
With SPF mailing lists continue to work as is, and forwarders
wishing to support SPF only need to modify the SMTP MAIL FROM
in addition to the RCPT TO, not the mail. And that's no unheard
of new concept, with the original RFC 821 SMTP forwarder always
added their host name to the reverse path in the MAIL FROM.
The most problematic point in the core SenderID specification
is its recommendation to interpret <tt>v=spf1</tt> policies like
<tt>spf2.0/mfrom,pra</tt> instead of <tt>spf2.0/mfrom</tt>.
This was never intended by all published SPF drafts since 2003,
and for an unknown large number of <tt>v=spf1</tt> policies an
evaluation for ''pra'' could cause bogus results for many cases
where ''pra'' and ''mfrom'' are different.
This technical problem - in fact only four letters <tt>,pra</tt>
in the core SenderID specification - caused the eighth appeal to
the Internet Architecture Board (IAB) in the Internet history.
In response to a prior appeal the [[IESG]] already noted that
SenderID cannot advance on the [[IETF]] standards track without
addressing the incompatibility with a MUST in RFC 2822.


==See also==
==See also==