lindberg@cs.chalmers.se (Gunnar Lindberg) (03/17/89)
We've run into an interesting problem: Somebody got a mail from a user at "edvz.uni-salzburg.ptt.at" and tried to reply to it. Note the top level domain, which is registered for Austria with "uunet.uu.net" and "aos.brl.mil" as NS:s (MX only). Now, what happened was that we tried to forward it and thus re-wrote it using "%...@"-syntax (yes I know, it's not "pure RFC822" :-) into: "user%edvz.uni-salzburg.ptt.at@kth.se" Then "sendmail" found the "at" and changed it into an "@" - look at "sendmail.cf", rule S3: R$+ at $+ $1@$2 "at" -> "@" for RFC 822 Guess what happened to "usr%edvz.uni-salzburg.ptt.@@kth.se"... Those who just said "it's non RFC822" should try "usr@foo.at.bar", :-). I've tried to just exclude the line from "sendmail.cf" and it seems to work quite well since "sendmail-5.59" catches "user at domain" in its argument list and converts it there. What doesn't work is the "To:" and "Cc:" lines, which will then stay as "usr at domain". I also guess "-t" doesn't work any more, although I haven't tried it. Has anybody else seen this? Is there a simple fix to it? Gunnar Lindberg
lindberg@cs.chalmers.se (Gunnar Lindberg) (03/17/89)
In article <2898@fnatte.cs.chalmers.se> I wrote about problems with
"sendmail" parsing "foo.at.bar" (RFC822 says "at" -> "@" :-).
The problem comes from rule S3:
R$+ at $+ $1@$2 "at" -> "@" for RFC 822
Now, I think I've found a fix:
+ R$*at.$* $1<@>.$2 save at.foo.bar
+ R$*.at$* $1.<@>$2 save foo.at.bar
R$+ at $+ $1@$2 "at" -> "@" for RFC 822
+ R$*<@>$* $1at$2 restore at.foo.bar/foo.at.bar
Gunnar Lindberg
gandalf@csli.STANFORD.EDU (Juergen Wagner) (03/18/89)
I do not see why the 'at' => '@' hack is necessary at all. RFC822 does not mention anything of that sort. In fact, sendmail.cf as distributed with 5.61 does not show a rule like the one you mention. -- Juergen Wagner gandalf@csli.stanford.edu wagner@arisia.xerox.com