Re: The Fact of relational algebra (was Re: Clean Object Class Design -- What is it?)

From: Patrick Logan <patrickdlogan_at_home.com>
Date: Mon, 10 Sep 2001 00:38:44 GMT
Message-ID: <oYTm7.58677$MK5.32446025@news1.sttln1.wa.home.com>

<D_at_B.A> wrote
in message news:XHSm7.1515$%u4.1538_at_www.newsranger.com...

> When insering into a collection does OODBMS locks the whole collection? Or
> programmer explicitly specifies the lock?

In Gemstone you have the choice. A "reduce conflict" collection (in the Smalltalk
database) or a "concurrent update" collection (in the Java database, same think,
different terminology) will support multiple sessions concurrently adding to the
same persistent collection.

"Regular" persistent collections can be locked. Otherwise concurrent update will
cause all but the first session to fail to commit.

The problem with concurrent update gets really bad with indexes, which are not
concurrent. So you really need locks with indexing in Gemstone.

> When object people don't even have a common object
> definition, how can they hope to be able to handle hard subjects like
> concurrency, for example?

It is a real problem. Just one reason why there is a risk in choosing a significant
investment in an OODBMS.

> Of course, this issue doesn't matter for CAD systems, but I wonder are
there
> crazy people who implemented a banking system with OODBMS?

Yes. I wouldn't call them "crazy". There are advantages and disadvantages. They have to be weighed using the specific circumstances.

--
Patrick Logan
mailto:patrickdlogan_at_home.com

Received on Sun Sep 09 2001 - 19:38:44 CDT