I've encountered the issue described here: http://mail.zope.org/pipermail/zope/2005-August/161120.html on a recently upgraded Zope using LocalFS. I have tracked down the issue to the fact that the object that LocalFS "hands" to ZODB is a WrapperObject, and this object does not have an oid. This raises an AssertionError during the _commit method of ZODB.Connection. I have traced the code a little bit and I see that this block of code will be avoided for these objects if the objects do not appear in self._registered_objects, which I can achieve by making sure that the object has no _p_jar attribute. If I comment out the line in the LocalFS object where the wrapper object's _p_jar is set to the parent object's _p_jar, everything works fine, with no (apparent) problems. [ I tried also setting obj._p_oid = self._p_oid, but then Zope has pickling problems trying to pickle instances of the wrapper class ]. The problem I have with my "solution" is that I have no clue what _p_jar is. What is _p_jar used for within the Zope model? What are the implications of having an object with self._p_jar=None? What are the implications of a transient object (not stored in the ZODB but on the local filesystem), with self._p_jar=None? Thanks in advance for any information, John Ziniti