Message formats

From: Eric Fischer <eric_at_fudge.uchicago.edu>
Date: Thu Jan 7 14:56:04 1999

> From: "Sean 'Captain Napalm' Conner" <spc_at_armigeron.com>
>
> > And while we're at it, why don't we complain about a single '.' on a line,
> > or starting a line with "From" :-)
>
> You can thank Eric Allman for those blunders. RFC822 specifically state
> that the SMTP server is to NOT modify the email as it goes through (as long
> as it conforms to the standard US 7-bit ASCII character set. Anything
> outside of the 7-bit ASCII character set is implementation dependant and
> anything goes). But in order to remain compatible with existing Unix
> software, lines starting with 'From ' had to be ``quoted'' else mail reading
> agents would get horribly confused

You can't really blame Eric Allman and sendmail for that -- the mbox
format was already established years before sendmail (or RFC822) was
written. And it *was* a reasonable format in the early versions of Unix,
when the mail command just printed out the entire mail file rather than
having commands to deal with individual messages. It should have been
fixed in version 7, but it wasn't, so we're still stuck with it.

Even so, it's not sendmail that does the quoting, unless you specifically
configure it to do so. My mail all goes through sendmail, but since it
goes directly (via procmail) into MH-format folders, lines beginning
with From make it through without getting mangled.

> (why they stored email in a single file is anybody's guess. I assume
> because early Unix systems couldn't deal with dozens or hundreds of
> files in a directory).

I imagine it was for efficiency of storage, assuming it had occurred
to anyone to even try delivering mail to multiple files. When you have
small disks and lots of users, you don't want to waste an inode and
an average of half a block of disk space on every single message.

Eric (not Allman)
Received on Thu Jan 07 1999 - 14:56:04 GMT