Re: Gnus, movemail, POP3, trailing empty lines

[Bernard.Steiner@Germany.EU.net]

     The POP3 server spits out a message at a time.  The last line of some
     messages is empty, or some others it is non-empty.  movemail concats
     all messages.  This results in a file that contains several messages,
     one after the other, and in some cases, a message begins with
     "\r\n\r\nFrom " (twice CRLF) whereas in other cases, a message begins
     with "\r\nFrom " (just one CRLF).  Gnus tries to read this file and
     does something strange when it sees the just-one-CRLF case.

Not exactly. You should be aware of the fact that all occourances (sp ?) of
"\r\n\r\nFrom " result from a non-initial mail in the mailbox, where the
preceding mail happens to contain a trailing empty line.
If you then interpret the empty trailing line (with the "\r\nFrom ") as a
separator, this may turn out OK. However, there's no reason why a mail should
contain a trailing empty line, and such a thing is certainly not required by
RFC 1939 (quoted below)

   Responses to certain commands are multi-line.  In these cases, which
   are clearly indicated below, after sending the first line of the
   response and a CRLF, any additional lines are sent, each terminated
   by a CRLF pair.  When all lines of the response have been sent, a
   final line is sent, consisting of a termination octet (decimal code
   046, ".") and a CRLF pair.

BTW the whole business about multiple mails with separators utterly fails
to produce recoverable results: Just consider PGP-signed mails that contain
an empty line, followed by a single "From " such as in

>From what I can see, this is a legal string iside a mail.

Now, you can say that your clever little mail-sorer will escape that to yield
a ">From " line, but then, the unpacking side will (correct me if I'm wrong)
never do a delete-one-greater-sign-from-a-string-matching-"^>>>*From ".
This is likely to produce checksum failures.
     
     I do not know whether a POP3 server should be required to tack on an
     extra CRLF to the end of each message, thereby ensuring the twice-CRLF
     case always, or whether movemail should tack on the extra CRLF, or
     whether Gnus should be taught to do without the extra CRLF.

     But somehow I think that the POP3 server is *not* the problem here.

Nope, it just is required to make sure there's a \r\n.\r\n sequence to signify
end of mail, where the CRLF pair before the dot belongs to the message.
See RFC 1939.

Bernard