[ZODB-Dev] Stability of ZODB in the Zope-3x-branch

[Jeremy Hylton]

On Thu, 21 Mar 2002 11:01:08 -0500
 Neil Schemenauer <nas@mems-exchange.org> wrote:
> We would really like to starting using Python 2.2
> features and get rid
> of ugly ExtensionClass hacks.  What's the stability of
> ZODB in the
> Zope-3x-branch?  Is it possible to start using it now?
> If not, what
> needs to be done?

Lots of testing -- more unit tests, user tests, etc.

A lot of the code is new, and it is based on two things I
only partly understood -- the new 2.2 features and the
BTrees implementation.  We've been finding little bugs as we
go along, and I'm aware of some other current bugs.  Ex:
There's a race condition in cache invalidation in
ZODB.Connection.

So I think the existing code needs careful review.  We need
better tests and we need users to bang on it.

There are also important architectural changes planned, so
at times the code base will be less stable that it is now.
Examples: Multi-version concurrency control, more
flexibility with isolation levels (kind of goes together
with MVCC), improved Cache (likely based on Toby's), etc.

I wouldn't put it into production use yet.

Are you interested in contributing on any of this work?
Perhaps we could have a ZODB4 sprint...

Jeremy