On Thu, Mar 03, 2005 at 05:00:17AM -0500, David Binger wrote:
> This risk could be avoided, I think,  by forcing copy_obj to actually
> close the exporting connection (or at least to ghostify every loaded
> object).

With a few interface changes, ghostifying can be put into the
export_object() function, so that shouldn't be a problem.  It does
mean that copy_obj() is somewhat

Incidentally, to jump up a level... all this effort wouldn't be
necessary if it was possible to conventionally pickle Persistent
subclasses.  Should I expect pickling to work?  If not, what could I
do to make it work?

--amk