[SCM] Samba Shared Repository - branch v3-2-test updated - initial-v3-2-unstable-713-g7a911b3

[Rafal Szczesniak]

Volker,

On Tue, Dec 18, 2007 at 12:27:32PM +0100, Volker Lendecke wrote:
> The reasons why I chose not to use memory-based tdbs for this
> data structure are different ones though. First I found it
> simply a bit silly to implement our own memory management
> when malloc is around and hopefully heavily optimized.
> Second, even after thinking about this for really long I
> could not find a way to implement something close to a LRU
> scheme inside tdb. The problem is that I can't really use
> pointers, tdb shuffles data around. So within tdb I would
> have to work with tdb keys to fake a doubly linked list and
> this would have involved much more memory usage in
> particular for the statcache where the keys are typically
> longer than the 4 or 8 bytes for a pointer.

Agreed. I didn't propose using memory-based tdb. It would not
allow you the required degree of control.

> But I would be happy to see a common API between the two,
> potentially based upon something similar to the dbwrap API
> with function pointers off the cache struct.

That's what I was actually thinking about. Common API with different,
let's say "backend", for data storage.

> Are you willing to do this?

Well, I could. However, I'd have to know more about how do you
want to operate on cache entries to be certain that current API
is enough to support both caches.


cheers,
-- 
Rafal Szczesniak
Samba Team member   http://www.samba.org
Likewise Software   http://www.likewisesoftware.com

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://lists.samba.org/archive/samba-technical/attachments/20071218/5a2211fd/attachment.bin