Development Discussion Board

I consider that pollution of the namespace... I see no changes in how we structure the DBAL in 3.2 from what we have right now

Having ported two more drivers over _today_, I see nothing wrong with how we do it right now... Perhaps if you could give a more "full" proposal?

The style the user has can differ from the default style. Him being logged in might have a different style than him being logged out. However, work is underway to increase the efficiency of the style.php script.

In the end, functionality wins. Without an alternative implemented here and now, {S_FORM_TOKEN} is here to stay...

The posting system is getting a shot in the arm, we will have soft deletes.

I highly discourage using CVS on a live board at this juncture.

In my post, there are examples where using a LEFT JOIN gives 20% speedup on an INNER JOIN. The advantage may be better given context change, memory use, CPU time, etc .. You can look at this post which is the perfect example : http://area51.phpbb.com/phpBB/viewtopic.php?f=3&t=29260&start=40...

Right, all the transaction code becomes a NOP if the underlying engine does not support transactions. phpBB installs the schema using the default installed engine.

In this case, it looks like the query planner chose the same join order on the tables...

This is too slow: 2008-01-04 21:53:58 CET LOG: instruction : SELECT u.username, u.username_clean, u.user_id, u.user_type, u.user_allow_viewonline, u.user_colour, s.session_ip, s.session_viewonline FROM phpbb_users u, phpbb_sessions s WHERE s.session_time >= 1199479738 AND u.user_id = s.session_user...

That one is quite slow: 2008-01-04 19:31:15 CET LOG: instruction : SELECT g.group_id, g.group_name, g.group_colour, g.group_type FROM phpbb_groups g LEFT JOIN phpbb_user_group ug ON ( g.group_id = ug.group_id AND ug.user_id = 1 AND ug.user_pending = 0 ) WHERE g.group_legend = 1 AND (g.group_type <>...