Persistence: a proposal

[Scott L. Burson]

   From: Patrick Logan <patrickl@servio.gemstone.com>
   Date: Thu, 22 May 97 23:20 PDT

   It sounds as if it is aiming at an area where Gemstone is weak.

This seems to be a fundamental tradeoff -- if you do things on an object basis
you have good concurrency and efficient small transactions, but it's hard to
efficiently support processes that create or modify 10^6 objects at a shot.
If you use some larger grouping you can handle big updates well, but the
concurrency and small-transaction properties become so poor that, in the case
of ZetaBase, one can only barely call the resulting thing a database.  (A
friend of mine coined the name "ZetaBase" and the rhyme was too good to pass
up, but really it's more accurately called a persistent object store than an
OODB.)

That said, I think there is actually some hope for making ZetaBase better at
small transactions than it is now.  I was working on this at one point, but
didn't get it completely worked out, and don't recall how close I got.  I have
some notes on it around here somewhere...

Anyway, yeah, it seems that there are a variety of ways to build an OODB, but
none of them (at least, none of those presently known) does everything well.

But I eagerly await hearing more from Kelly about his/her POS/OODB.  (I know
there aren't a lot of female Lisp hackers, but I wouldn't want to jump to
conclusions, and the name "Kelly" provides no clues.)

-- Scott