[ZODB-Dev] cache not minimized at transaction boundaries?

[Tim Peters]

[Chris Withers]
> ...
> ...oh well, if only the ZODB cache was RAM-usage-based ratehr than object
> count based ;-)

Ya, that's not feasible.  More plausible would be to base ZODB cache targets
on aggregate pickle size; ZODB _could_ know that, and then it would also be
strongly related to how a ZEO cache's size is measured (the ZEO cache size
limit is given in bytes, and by far the bulk of those bytes are consumed by
pickles).