Why use messageblks? Well, for one reason, mysql has a limit on the record size so a split-up is necessary anyway. The system was chosen to handle big messages in chunks.
----- Original Message ----- From: "Sam Przyswa" <[EMAIL PROTECTED]> To: <dbmail@dbmail.org> Sent: Tuesday, September 24, 2002 3:52 PM Subject: RE: [Dbmail] Imapd is slow as hell > Micah Stevens ([EMAIL PROTECTED]) wrote: > > > >Just a note, on your query comments, although I'm not arguing with your > >basic point, it's always been my understanding that SELECT statements that > >use '*' instead of stating field names are slower. If you don't specify the > >field names the database has to spend time figuring out what they are, but > >if you name them all it has to do it error if it can't find the name. > > > >In practice I've noticed it to be only marginally faster to explicitly name > >the fields rather than use the '*' wildcard. Not really noticeable unless > >you're doing tons of SELECT statements. > > Either or not use wildcards it should be faster to get the header in one query > instead of five, I think. > > >Back to the point of your mail though: Is it faster to make one big > >conglomerate SQL statement that takes the database longer to parse? If it's > >the difference between one connection, and two, I doubt it, but the more > >you slim things down the better. That's for sure. > > If the message is stored directly in "messages" instead of "messages" > + "messageblks" the query look like: > > "SELECT * FROM messages WHERE message_idnr = <msg_nbr> AND status<2 AND unique_id ! > = '' AND mailbox_idnr = <mbox>" > > ...or for exemple: > > "SELECT from,to,subject,seen_flag FROM messages WHERE message_idnr = <msg_nbr> AND > status<2 AND unique_id != '' AND mailbox_idnr = <mbox>" > > My second question: why to use the extra table messageblks to store the message ? > > Thanks for spend your time to reply to my questions. > > Sam. > > >At 03:26 AM 9/24/2002 +0200, you wrote: > >>Hi, > >> > >>We have noticed that and make some tests with several Webmail and mail > >>clients as Netscape, Outlook, vs courier-imap with the same mail > >>clients. > >> > >>First conclusions the PHPs' webmails are very slow except Squirrelmail > >>because it use its own imap libraries instead of PHP-IMAP. But it's true > >>Courier-IMAP it's much faster than dbmail-imap, we have add the mysql > >>indexes as explained in this list, tuned the mysql config and got some > >>improvements but never too fast as courier-imap. > >> > >>Then we have put the trace_level to 5 and make the same tests and we > >>have noticed that to get each mail header dbmail-imap need 4 or 5 > >>queries and in my opinion I think that it should be possible to get the > >>same think with only one or two: > >> > >>By use a "SELECT * FROM messages" instead "SELECT field-1, field-2..." > >>and/or change the table structure to put the mail header on "messages" > >>whith the from:, to:, subject:, field and perhaps the body too, or just > >>the body on messageblks. > >> > >>DBmail is the right way to make an isp mail solution with a lot of > >>capabilities as CRM, etc, but we have to improve its speed and discuss > >>to find the best way to do. > >> > -- > Sam Przyswa - Chef de projet > Arial Concept - Intégrateur Internet > 36, rue de Turin - 75008 - Paris > Tel: 01 40 54 86 04 - Fax: 01 40 54 83 01 > Web: http://www.arial-concept.com - Email: [EMAIL PROTECTED] > > > _______________________________________________ > Dbmail mailing list > Dbmail@dbmail.org > https://mailman.fastxs.nl/mailman/listinfo/dbmail >
