[ZODB-Dev] Multi-version concurrency control

[Jeremy Hylton]

On Wed, 2003-10-08 at 11:27, Steve Alexander wrote:
> >>Why not extend the proposal to remove versions from ZODB 4 altogether?
> > 
> > 
> > There are a lot of reasons not to remove versions.  We don't have any
> > plans to work on ZODB4 at the moment, so it wouldn't get done for a long
> > time anyway.
> 
> When I read the proposal, I assumed that the MVCC work would be on ZODB 
> 4, not ZODB 3.
> 
> I guess that was a combination of my mind hanging on to previous 
> discussions about the roadmap for MVCC, and that the proposal doesn't 
> say what version of the ZODB it applies to.
> 
> I have a small team working on an application that uses ZODB 4. We 
> expect it to be used by some real people in 2 - 3 months. Should we 
> consider changing to ZODB 3?

Yes.  It's a complicated question, though, because ZODB3 still requires
ExtensionClass.  We've got an experimental branch that removes
ExtensionClass, but I don't know what the schedule is for getting that
into a mainstream release.  The MVCC code will probably get done first.
We don't have persistent classes in ZODB3 yet, either.

On the other hand, there's no active development going on for ZODB4 and
a growing number of bugs fixed in ZODB3 and not ZODB4.  I wouldn't trust
the ZEO code in ZODB4, for example.

Jeremy