I use sendmail's virtusertable feature to deliver mail for
[EMAIL PROTECTED] to somelocaluser; I was thinking of modifying the Cyrus
code that handles IMAP LOGIN so that it tries to look up the username as
presented by the user in the sendmail virtusertable, and if it succeeds,
it will log in as the mapped name, so the user never has to know the
somelocaluser part. This will do the job for me, but I can see some
problems with it:
- Need to make sure that Cyrus can locate and read the virtusertable map,
which could mean several things, including making sure sendmail and Cyrus
are built with the same BerkeleyDB. If virtusertable is elsewhere (LDAP
routing?), then something else has to happen, and I'm not at all sure how
to express this from a configuration perspective, let alone
implementation.
- Public folders would become difficult for users, because they'd have to
have a way to learn their somelocaluser
I'm sure there are more problems.
Meanwhile, this seems to me to be a way to implement virtual domains in
such a way that we don't have to agonize over Cyrus path separators,
sendmail hacks, multiple IP addresses, and making lmtpd less enthusiastic.
On balance, is this a less invasive approach than all of the other hacks?
Pete.
