Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!lll-winken!uunet!ibmarc!ks.almaden.ibm.com!kurt
From: kurt@ks.almaden.ibm.com (Kurt Shoens)
Newsgroups: comp.mail.misc
Subject: Re: Screwey 'r'eply addresses
Summary: UCB Mail thinks .'s are significant
Keywords: UCB-mail reply mixup
Message-ID: <740@ks.UUCP>
Date: 23 Mar 89 15:59:55 GMT
References: <1616@thor.acc.stolaf.edu>
Sender: news@ibmarc.UUCP
Reply-To: kurt@ibmarc.UUCP (Kurt Shoens)
Organization: IBM Almaden Research Center, San Jose
Lines: 43

In article <1616@thor.acc.stolaf.edu> fritchie@thor.stolaf.edu (Scott Fritchie) writes:
>Ever since we joined the Internet, we've been having some problems with
>using the 'r' reply command in with UCB mail.  Occasionally, the addresses
>the reply comes up with are horribly scrambled.  Here is a sample message:

Scott goes on to give an example replying to a message with...
   From fritchie@agnes.acc.stolaf.edu Sat Mar 18 15:13:14 1989
   To: fritchie
   (etc and so on)

Replying to this yields
   To: agnes.stolaf:fritchie@edu fritchie@agnes.acc.stolaf.edu

I think the phrase "every since we joined the internet" is the key clue.
Mail's trying to be clever and tack the "from" hostname onto the to:
addressee "fritchie." From looking at the code in mail (that horrible
dreck in optim.c) it seems that .'s are considered by mail to separate
significant parts of addresses (i.e., parts that mail should be fooling
with) and for domain style names, this is untrue.  In the file config.c,
there is an initialization of the string "metanet" which might be improved
by removing the . from the list.  Also, the reason that :'s drift into the
address has to do with the way mail preprocesses addresses with @'s in them.

In the name of thoroughness, I just tried this on my own copy of Mail
and duplicated the symptoms you described.  The change to the definition
of metanet in config.c produced the correct rewriting.

The R command doesn't alter the address at all; it just takes the address
that the message is from.  That's why you're not seeing any scrambling of
those addresses.

The rewriting of addresses by Mail as a Larger Issue was debated briefly
in this space a few weeks ago.  Some folks argue that you will be better
off if you cause sendmail to rewrite the To: and Cc: addresses so that
they are fully qualified.  In fact, the addresses have to be rewritten
this way to conform to RFC822 (and they new one, whose number I don't know)
anyway.  Then, rewriting of addresses on reply is not needed.  I think
that MH requires the addresses to be fully qualified for replies to work.
The advantages of having sendmail do the rewriting is that 1) sendmail's
mechanisms to rewrite are very flexible; and 2) sendmail's rewriting can
be changed without recompilation.

Kurt Shoens, IBM Almaden Research Center, SHOENS at ALMADEN