On Mon, Jul 21, 2003 at 10:42:45PM -0400, Eric S. Raymond wrote:
Rob MacGregor <rob_macgregor(_at_)hotmail(_dot_)com>:
The general case is nastier than you know. I wish *I* didn't know.
Even with those cases, I can see some value in Michael's suggestion that
corrupt/mangled emails be passed, as an attachment, to the postmaster (or
possibly to a "corrupt email" address, that way sites that only see this in
spam can black hole it). Any chance of having such a feature added?
If nothing else, it might stop the regular flood of emails to the list
about the problem :-)
Well, I cleaned up some stuff in 6.2.2 that was contributing to the problem.
Mail with bad headers is no longer retained on the server; instead. we try
to deduce a location for the header/body split.
Let's see what corner cases are still causing problems before we try
anything more elaborate.
I'm not sure if I have a edge case or not.
Last time I got some of this malformed email (under 6.2.3), it hit a
piece that was malformed and said so. Then I got a protocol error.
Errors looked like this:
fetchmail: message delimiter found while scanning headers
flushed
fetchmail: client/server protocol error while fetching from <pop server>
fetchmail: Query status=4 (PROTOCOL)
There were more messages on that server, but it apparently broke the
connection at that point. When I went back to fetch the rest, _all_ the
mail was still on the server. (I had to run fetchmail -F to kill it.)
I've never had a protocol error with this server, and I haven't had any
malformed-header emails since, so I can't say whether the behavior is
consistent. (I'll let you know if it happens again.)
If others have seen this behavior in 6.2.3, or if the phenomenon rings a
bell with you, then maybe it represents an edge case.
Paul