Re: Db2, Oracle, SQL Server

Serge Rielau wrote:

> > I have already advised your sorry lot to get a copy of Steve Adams'
> > book and LEARN how locks should be done properly in a database.
> I don't know this guy.. But his resume certainly implies a certain
bias
> which may explain why he's not that widely known beyond this
crowd....

Forget the resume. His book contains the best explanation you can find anywhere, including Oracle docos, of how Oracle row locking really works and why it doesn't need any of the rigmarole of memory/disk/whatever IBM and others keep attributing to it.

If IBM really truly wants to criticize the locking mechanism of Oracle (and I'm the first one to say it should do so), then it is quite fitting that you folks LEARN what it is all about. And Steve's book is the only reference out there that completely explains all those things.
You may take the pedantic approach of ignoring it, but it will be at the cost of credibility of any IBM criticism, believe me.

> If we look at a random, more generic and most certainly less biased
CS
> lecture (found on google with 30sec investment):
> http://www.cse.mrt.ac.lk/pg/msc/lecnotes/cs5251/Lecture8.ppt
> I find little stating that multiversion concurrency is "better".

It's got NOTHING to do with multi-versioning. We are talking row locks on updates. Can you folks stay on the subject for once?

> Its just a different techique, just like e.g. optimistic locking.
> Each has its uses, up and downsides.

I find no downsides in Oracle's row locking. Multi-versioning may have some, but that is separate from true row locking. Of course, you may think that a mechanism that implies and needs lock escalation is superior. I don't think so and there are a lot of people out there who share my opinion.

> Isn't it ironic that if SQL Server 2005 ever ships it will be the
only
> mainstream DBMS supporting all major isolation levels?

At what cost, though. And: are they really needed? And: isolation levels are not exactly the same as update locks. Too many imponderables. But yes, it is ironic.

There is one thing about SS that is truly endearing it to me: its ability to use any language for SPs. Not just T-SQL. That is a great idea and it should be adopted by others. And before anyone mentions the bleeding obvious: no, Java is a piss-poor language for data manipulation and I hate having to be stuck with it in Oracle if I want to use aything else than PL/SQL. Received on Tue Feb 08 2005 - 01:01:29 CST