ODMG

On Thu, 12 Oct 2000, Aristide Aragon wrote:

I was wondering...
Are there any plans to include ODMG support for psotgresql? ...
Or does anybody know of an open source ODMG DBMS?
It'd be cool if besides being object-relational, postgresql were
object oriented too!

From my understanding, based on a long, and informative, conversation with
Thomas last week, it doesn't sound like its possible to be both OR *AND*
OO at the same time. We can implement some OO features, but, at least the
way I understood it, to go fully OO would mean removing the whole OR ...

If you, and/or a group of ppl, want to work on OO features that can be
merged in *without* affecting the OR base, plesae feel free to do so
... we even have a mailng list (pgsql-hackers-oo) setup for discussing
this issue, but other then a few threads when it was first setup, it kinda
grew dead :(

Besides, OO has it's moments, but I've yet to be convinced that it's a
viable step in evolution. Apart from high-speed OID implementations, I'm
not sure that they gain that much. Certainly, there are no features like
stored procedures, views, rules, etc. and the query interface is
rudimentary, and immature. Of course, maybe I'm tarring a lot of good
products because of my limited experience, but I've had a reasonable look at
two supposed market-leaders, and I'm completely underwhelmed. The only way
to manipulate data is to instantiate it, which is outside of the database
process. And besides, I'm still waiting to see an OODB that can index a
complex number ;-) or an IP address.

I like the idea of OR databases, the problem thus far is that, apart from
PG, there are no good implementations of OR. All the commercial
implementations are just plain poor. They may perform certain functions
well, but they are not complete, and that's no good.

I suspect the time is near for PG to worry less about relational adherence,
and start leading the OR drive. It's positioned better than any other
player in the market. Not to say that relational is not important, it's
still most of the market, but that market is shifting, and OO databases are
only niche products. There is a whole huge market out there which really
needs OR, there is only one real contender, and that contender is more
focused on basic relational functionality. Of course, PG really needed
things like DRI, and there are still some things that are required before
the team can say "Been there, done that, got the T-shirt", but perhaps
starting to think about OR positioning wouldn't hurt.

Perhaps a good OID implementation would be a good idea, but apart from that,
the rest of OODB is fancy footwork, which I'm not sure would significantly
benefit us, without us losing a whole bundle.

Cheers...

MikeA