Ian Bicking wrote:
On Tue, 2003-01-28 at 20:29, Stuart Donaldson wrote:
** Regarding the SessionMemoryStore and saving to disk. **

It appears unlikely from the e-mail thread that many people are using 
the SessionMemoryStore relying on the save to disk behavior.  The 
direction I think we are leaning is to remove that capability of the 
SessionMemoryStore, making it memory-only and not persistent.

I don't think it's a good idea to change this, at least not at this
point.  It seems better to create a new session store
(NonPersistentMemory or something), and perhaps to deprecate Memory
(i.e., print a warning message).  Then maybe restoring a new Memory
behavior (i.e, reusing that name), or just adding another kind of
session store.

I think it would be fine to include such a new session store
(particularly if someone offers tested code for it), since it shouldn't
effect the rest of Webware very much.

Agreed that this is not the time to change the behavior of SessionMemoryStore.

I think what would be useful, is a mechanism to have non-persistent data associated with a session.  But to not necessarily require that all session data be non-persistent.  This can be done now via a mechanism similar to what I described earlier, using a non-persistent dictionary keyed on the session, and containing the non-persistent data.  However it would be nice to have a "standard" API for this at some point.

But I digress, at this point I think even for that, it would be best to address it after the 0.8 release.  Putting something in the release adds pressure for some level of support in the future.  And if we are adding new functionality to the release, I think it better to have it thought out and tried for a while in development, to give others more opportunity to comment on it and refine it.  Just my 2-bits.

If someone wants to propose a NonPersistentMemory store as an experimental feature, I could see including it.  But I want to release 0.8b2 on Wednesday, and I would like ot to be a final beta prior to release.