yeah it is about time we start thinking about it ..
top-issue: shared folders, imap performance
shared folders could be implemented quite easily in dbmail 1.1 - if you
know your way around mysql you can create a (bit unstable) shared
folder yourself. Main issue is to change the UNIQUE(mailbox_idnr) to
UNIQUE(mailbox_idnr, owner_idnr). This way mailboxes created using the
same idnr but different useridnr's are shared. Imap code needs to be
changed a little (we cannot longer trust queries using just
mailbox_idnr as identifier for a single mailbox) but these are, in
principle, minor changes.
as for the imap-performance:
first of all, the number of queries needs to be reduced. A smarter
parsing of the commandline in a FETCH command could save _lots_ of
queries.
moreover, my plan is to save parsed message-info into the database.
This would put some more strain on the smtp inserter but as it is
obliged to read the entire message one could as well parse it. The
imapserver would be very, very much faster. I have thought this over
now a little and am hoping to send you all some ideas about
implementing this in terms of table structure and program logic.
Finally, i would like to create an abstraction layer such that messages
can be linked to multiple mailboxes. The win here is not so much
benefits for the smtp but again to the imap server as the imap protocol
does not have a MOVE command - when you move a message in your
mailclient the mail is now copied then deleted requiring all the
messageblocks to be re-inserted. With this new layer a COPY would just
mean inserting a link-entry occupying just a few tens of bytes instead
of kilo- or even megabytes.
These are the plans so far, please consider these ideas and give me
your thoughts about it! It would be great to have a plan for dbmail 2.0
by the end of next week.
regards roel
_________________________
R.A. Rozendaal
IC&S
T: +31 30 2322878
F: +31 30 2322305
www.ic-s.nl
