Date: 8 Jun 1999 04:34:52 -0000
Message-ID: <19990608043452.2965.qmail@cr.yp.to>
From: "D. J. Bernstein" <djb@cr.yp.to>
To: ietf@ietf.org
Cc: drums@cs.utk.edu
Subject: Re: Asinine anti-spam mechanisms
References: <19990606030904.21624.qmail@cr.yp.to> <199906060337.XAA09500@astro.cs.utk.edu> <19990606060920.22356.qmail@cr.yp.to> <199906070202.WAA04331@saint.heaven.net> <19990607213502.1218.qmail@cr.yp.to> <199906072345.d57Njp029878@black-ice.cc.vt.edu>

See http://pobox.com/~djb/docs/copies.html if you're curious about the
bounce message that has gotten Kletnieks so upset.

Valdis.Kletnieks@vt.edu writes:
> RFC1891 through RFC1894 are for building less fragile mail systems,

That's what they were _intended_ to do, but intent isn't enough when it
isn't backed up by solid engineering.

RFC 1891 is an absurdly complicated link-level protocol, unsupported by
most MTA authors and useless for standard multidrop POP mailboxes. Its
misguided notification guarantee, which is what forced the complexity in
the first place, appears to have been destroyed by sendmail.

RFC 1892 and RFC 1894 aren't as ludicrous as RFC 1891, but they're still
unnecessarily difficult to parse.

Fortunately, all of these RFCs have been rendered obsolete. The big
problem that motivated them---namely, the difficulty of identifying the
subscriber for each mailing-list bounce---has been eliminated by VERPs.

> 'least likely to be machine parseable'

In fact, qmail's bounce messages follow a documented, easy-to-parse
format that has around just as long as RFC 1894. Several independent
parsers have been deployed. See http://pobox.com/~djb/proto/qsbmf.txt
for a description of the format and a comparison with 1892/1894.

---Dan