[ZODB-Dev] Packless, Mountable FileStorage?

[Barry A. Warsaw]

>>>>> "CW" == Chris Withers <chrisw@nipltd.com> writes:

    >>  Since the current implementation appends changes to the file,
    >> it would be a complete new principle to change this. I think
    >> using the BSDDBStorage gives exactly this without extra
    >> developement.

    CW> ...but introduces a dependency on BerkleyDB :-(

True.  And this means you accumulate all the stability and/or
headaches of BerkeleyDB.  For example, you will incur the
administrative overhead of dealing with BerkeleyDB transaction logs,
but you gain in reduced in-memory footprint.

    CW> for me, the big win with ZODB is it's "batteries included"
    CW> stuff. If I need a database to store the data then I might as
    CW> well go for something like PostGres or Oracle and be done with
    CW> it.

There's nothing preventing you from continuing to use FileStorage or
any other backend storage.  That's one of the advantages of ZODB's
architecture.  And while I agree about the "batteries included"
advantages, there are other advantages to ZODB than going directly to
the Postgres or Oracle layer, especially if your application is in
Python.  ZODB has probably the most natural and transparent match for
Python programs that have persistence requirements, while storing at
the SQL level involves lots of adaptation between Python and your
database.

    CW> I dunno what the ZC thinking is on this...

If it were me talking <wink>, I'd say it's all about tradeoffs, so you
need to decide what's best for your application.

-Barry