james at digisys
Apr 3, 2002, 10:58 AM
Post #1 of 1
(140 views)
Permalink
|
Hello Roel!
I completely agree about the fetch/store/copy comands, and they seem
pretty snappy in the CVS. Here is what I've been looking at. The one
command that IMAP clients (like Mozilla) do often when opening a mailbox
is "UID FETCH 1:* (FLAGS)". For what I see, this command goes into a
loop where a SELECT is done for each IMAP flag of each message in the
box. So my example mailbox (which is 4500 messages, a mailing list)
would be doing over 25,000 queries to return the results. (which pounds
the DB for like 11 seconds here) I was thinking that a much simpler
method for responding to "UID FETCH 1:* (FLAGS)" would be to do a single
query like:
SELECT message_idnr, seen_flag, answered_flag, deleted_flag,
flagged_flag, recent_flag, draft_flag FROM messages WHERE mailbox_idnr =
'X'
Then just parse that single set of query results. Seems this would be
increadibly fast in comparison. I realize that a UID FETCH command
could come in with ranges like:
"UID FETCH 418,499,512:550 (FLAGS)"
But even this could be done with 3 selects. Two for selects for the
individual messages and one for the message range. Basically 1 SELECT
for each element in the comma separated list.
I understand what you are saying about large mailboxes, many people do
use IMAP folders for things like mailing lists. And the beauty of
DBmail is that if the DB is properly indexed and queried it would have
the ability to handle very large mailboxes. Something that *no* other
imap server can claim. ;)
Hoping some of this makes sense. I'm still trying to get familiar with
the code, but this caught my eye to start. I'm still amazed at the
amount of work you guys have done on this. I was very glad to find it...
-James
|
