dovecot - Dec 2009 - mailboxcasecmp()

I remember way back, i'd modified the dspam training plugin to compare
mailbox names case-insensitively.  that patch wasn't accepted.  i didn't
understand why since it was obviously correct and had no negative
impact, at least i didn't think so.

Now I've hit another case where case-insensitive mailbox names would
be useful: +addressing.

If I have a mailbox 'Ext' and I send mail to frank+Ext@, the LHS
is lowercased to 'frank+ext' and the +extension delivery doesn't
find the 'Ext' mailbox.

If the +extension is going to be lowercased, then the mailbox names should
be lowercased when looking for them.  Or, since 'ext' and 'Ext'
can
co-exist, the +extension shouldn't be lowercased when looking for the
mailbox.

I prefer the former (lowercase everything), since it's unlikely a user
would create both 'mailbox' and 'Mailbox', however I can
envision plenty
of times where users would try to use '+ext' yet create the
'+Ext"
mailbox and never be able to figure out why it doesn't work.  I know the
latter solution (case sensitive extensions and case sensitive mailboxes)
is more correct but really it will be harder to debug user problems.

-frank

On Wed, 2009-12-02 at 02:19 -0800, Frank Cusack wrote:
> I remember way back, i'd modified the dspam training plugin to compare
> mailbox names case-insensitively.  that patch wasn't accepted.  i
didn't
> understand why since it was obviously correct and had no negative
> impact, at least i didn't think so.
I guess I didn't like that it was ascii-only. Nowadays Dovecot supports
i;unicode-casemap to do Unicode case insensitive checks. That could be
used.. But that could be done internally by dspam too. I'm not sure if
there's much point in adding mailboxcasecmp() that does exactly the same
as unicode_casecmp() (although a simple version of that doesn't exist
yet either).
> Now I've hit another case where case-insensitive mailbox names would
> be useful: +addressing.
> 
> If I have a mailbox 'Ext' and I send mail to frank+Ext@, the LHS
> is lowercased to 'frank+ext' and the +extension delivery
doesn't
> find the 'Ext' mailbox.
> 
> If the +extension is going to be lowercased, then the mailbox names should
> be lowercased when looking for them.  Or, since 'ext' and
'Ext' can
> co-exist, the +extension shouldn't be lowercased when looking for the
> mailbox.
Well.. I suppose the logic could be like:

1) Try opening the mailbox. If it exists, great, use it.

2) If it doesn't exist, try to find the mailbox case-insensitively.
There may be multiple levels, like user+work/foo/stuff at domain.org so it
would have to match all work, foo and stuff case-insensitively
separately. Mailbox listing also makes it slower, but if this isn't done
often it's probably not such a big deal.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
URL:
<http://dovecot.org/pipermail/dovecot/attachments/20091209/be4bcef8/attachment-0002.bin>