Arnt Gulbrandsen wrote:
I've tracked it down to buggy SMTP handling. Quoted-printable
encoding inserts a soft linefeed, after character n.
[...]
Yes, got it, and that triggered some dubious ideas: What if we
had a "bytes <nnnn>" clause in the timestamp lines (as an aid
to debug such problems) - in _theory_ it could work, substract
the length of header fields introduced by hop n+1, and then it
should match what hop n reported as length.
In _practice_ it's probably a bad idea, but with a side-effect
for the proposed "trace header field I-D": that text should
decree that anything adding trace header fields MUST identify
itself by the same "by" clause as in its timestamp line.
The obscure Received-SPF somehow managed to bypass this MUST
(cheap excuse, maybe I trusted that Bruce would kill it anyway
in an "IETF last call" before it escapes into the wild). Bye