On Sat, 2008-07-12 at 09:47 -0700, Mark Atwood wrote: > On Jul 12, 2008, at 3:26 AM, Antony T Curtis wrote: > > > > Here is my $0.02... > > > > Table/view/schema object definitions should all be persistent > > (forever). This makes it cheap to check if a table exists without > > having to hit the disk. An async event notification system should > > exist so that a storage engine can inform the server of the creation > > of a new table so that the server can request and cache it. > > The S3 storage engine doesnt know if an underlying table exists until > it tries to access it. New underlying tables could be created and > existing ones deleted at any time. S3 has no way of informing the > engine that a table has been created or deleted, thus the engine cant > inform the server either. > > The IBM DB2 SE has this same problem. > > Your idea has a lot of merit. But engines have to have a way of > saying "i cant tell you if a table exists until you try to look for > it, and you cant trust your cache to be utterly sure that the table > still exists".
Yep, and we can still end up halfway through executing the query before the engine throws a table doesn't exist error. As with NDB (and S3 and federated and any other remote engine) actually doing a synchronous check takes way too long.... and asynchronous checks are just.. well.. asynchronous. We must also remember that some installations have 20,000+ tables. Having those all, constantly loaded in memory kinda sucks. -- Stewart Smith ([EMAIL PROTECTED]) http://www.flamingspork.com/
signature.asc
Description: This is a digitally signed message part
_______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : [email protected] Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
