At 19:30 14/01/02, Gary Mills wrote: >On Mon, Jan 14, 2002 at 05:35:49PM +0000, John Holman wrote: > > > > In fact, even when the local part of addresses is case sensitive, I think > > it would be better to do a case insensitive match for purposes of deciding > > eligibility for a vacation message. After all, if the message is delivered > > (so the envelope recipient address is correct) and has a to/cc/bcc header > > that satisfies a case-insensitive match, under what circumstances could it > > be wrong to send a vacation message? > >A major part of this problem is that the envelope recipient address >is constructed in such a way that it can never match a header recipient >address. This happens because the MTA, sendmail in my case, strips >the domain portion of the envelope recipient, and then Cyrus lmtpd >appends `@unspecified.domain' to it. This requires a configuration >option so that lmtpd will append the correct domain. Lmtpd requires >some information only known to the MTA.
Agreed that lmtpd needs to know the email addresses to use when matching what appears in the to/bcc/cc headers, and cannot deduce those from the envelope recipient. In fact it uses information provided by the user in the vacation rule. I'm addressing a different issue: that the case sensitive comparison inhibits vacation messages in some cases where a case insensitive comparision would not, and that this inhibition is never (as far as I can see) what is wanted. Therefore a case insensitive comparison would be better - and should at least be a configuration option. > > If that is not agreed, perhaps there could be a configuration option for > > the vacation facility to do a case-insensitive match? > >Again, lmtpd needs to know some information, whether user names are >case-insensitive, known only to the MTA. It doesn't need to know that information if a case-insensitive match is always better. And if that is not agreed, the configuration option I mention would serve to give lmptd the information. >The real problem is that the Cyrus sieve module does not know the >e-mail address of the recipient, and must guess at it by by examining >headers. Solving this would solve the other problems. Even if Cyrus, or (perhaps more likely) an interface that generates the sieve vacation rule on behalf of the user, discovered the "personal email addresses" for somebody (e.g. via a directory lookup keyed on the username), addresses varying only in case should still be treated as equivalent when deciding whether to send a vacation message. And, if Cyrus made a case-insensitive comparison for this purpose, that would solve one part of the problem even if the user still had to specify their own email addresses when creating a vacation rule. John. >-- >-Gary Mills- -Unix Support- -U of M Academic Computing and Networking-
