[net.mail] New headers breaking mail

rjk@mgweed.UUCP (Randy King) (05/13/85)

<><><>

I'm not sure of the dynamics of this situation, but lately I have started
getting more and more mail from folks using some bizzare new mailer whose
headers are breaking my mail servers.  It a familiar problem; either the
mail gets tacked on to the previous mail or if it is the first mail in
/usr/mail then it is unreadable.  Following is an example of the header;
you will note that the date field in the From line is causing the problem.

	From mgweed!pegasus!hansen 12 May 1985  13:47 EDT
	FROM:       hansen@pegasus.ATT.UUCP
	TO:         mgweed!rjk
	DATE:       12 May 1985  13:47 EDT
	SUBJECT:    Re: ...

I really wish that people would not just start implementing their own
defacto mailers that break existing mailers.  I realize that this *may*
be just barely RFC standard, but surely these things can be introduced
in a more controlled manner.  People like Tony are the ones that suffer,
as their mail may not be getting through.

Would someone care to shed some light on this mailer?  Am I at fault?  Do
I need to fix my mailers?
						Randy King
						AT&T-CP@MG
						ihnp4!mgweed!rjk

honey@down.FUN (Peter Honeyman) (05/14/85)

You are surely not to blame here, Randy.

Your mailer, probably mailx, a derivative of Mail, wants
messages to be separated by a blank line followed by a valid
"From " line.  A valid "From " line has the following form:

From <user> <ctime format date string>

(Ignore any leading ">" inserted by rnews.)  <user> is any
string not containing whitespace.  <ctime format date string> is
exactly what it claims to be: the 26-character string described
on the ctime man page.

Locally, I validate the first line of each incoming message in
rmail.  If it's not right, I reject it immediately, with stderr
"No From line !?!"; uuxqt then takes care of the rest.  This has
always produced immediate results.  Emacs-based mailers seem to
be the most frequent offenders, since they tend to shun unix
mail altogether, and call uux directly.

	Peter