[Zope3-dev] Tossing Data.fs considered harmful

[Barry A. Warsaw]

>>>>> "GvR" == Guido van Rossum <guido@python.org> writes:

    GvR> Maybe as a compromise, we should provide conversion tools
    GvR> between the Data.fs flavors belonging to official releases,
    GvR> starting with the next alpha (I'm not sure if Jim's planned
    GvR> release should be considered an alpha or not; I've heard
    GvR> conflicting rumors about this).

I think Jim's calling the next release a "milestone" exactly because
some people think "alpha" is a strong line in the sand and won't
tolerate breakage across alpha releases.

I can't comment on application level stuff but from a ZODB standpoint,
we were planning on doing a ZODB4 milestone release to mirror the
Zope3 milestone release.  Knowing that there are ZODB changes coming
down the pike that will break the storages, I think it would be
reasonable to provide conversion scripts between milestones, if
possible.

I'm not sure this will be relevant but would you be willing to forgo
important ZODB4 changes at this early point if it were /not/ possible
to write a conversion script?

Also, we'll either have to write a generic conversion script against
the storage API or each storage will have to have its own conversion
script.  Then again, at least File and the Berkeley storages have a
storage version number so they could conceivably auto-update.

All that takes work though, so I would defer to Jim.  I think he
should decide about backward compatibility guarantees and timetables.
It may not make much sense to cement the storage formats if there are
no guarantees at the application layer.

-Barry