[Schevo-devel] Newbie questions
David Binger
dbinger at mems-exchange.org
Tue Oct 18 06:44:45 EDT 2005
>>> t applications with several
>>> objects that are edited frequently (Durus website says something
>>> against
>>> using it for these scenarios, unlike the ZODB which seems to handle
>>> things pretty well... expecially now with MVCC).
Durus and ZODB work pretty much the same way. Frequent writes to the
same object are not good in either case, or in any transactional
database
of any kind.
MVCC is clever, but misunderstood. First, it is not a magic bullet.
It just
allows operations to work in old transactions. This can be useful in
certain situations where you really don't care if the information being
used is actually transactionally up-to-date. I think most applications
really do care about transactions. Second, MVCC is not automatic.
The programmer must specifically add a _p_independent method to a class
that enables MVCC for instances of that class. Third, the Zope 3
source code has this line:
> Connection.py:755: # Does anything actually use
> _p_independent()? It would simplify
which suggests that, in practice, actual use of MVCC is not common.
More information about the Schevo-devel
mailing list