Going ZODB

[John Wiegley]

>>>>> On Sun Mar 11, Steven writes:

> I am curious as to how many people in this list actually use ZODB in
> production environment and what is the typical uses of it
> (especially ZODB with relational/berkeleydb storage). Aside from
> along with Zope, of course.

> Also, there are one issue I am considering before going ZODB. If
> later on the application grows beyond the scale that ZODB can
> provide at the moment, would it be awkwardly hard to convert the
> code to use other backend.  For example in the relational world, if
> I keep the SQL statements portable enough, I can just,
> theoretically, switch the backend with another DB.

I tried to use ZODB as the persistence engine for my MOO clone
(http://doop.sourceforge.net).  I found it's restrictions too onerous,
most especially the restriction that I not override __getattr__ or
__setattr__.  Lastly, I found that debugging problems was hideously
difficult, and that mostly I had to resort to a binary search using
print statements, just about all the time.  Maybe they've improved the
robustness of this.

Feature-wise, I would love to use it.  If I could ever find out how to
use ZODB *and* overload get/setattr, I think I'd switch.  What I'm
doing now (just checkpointing the whole database at time intervals) is
too basic.