Re: IBM Debunks Oracle's MultiVersion Read Consistency ?

Following up on Robert Allen, 21 Apr 2003:

> They raise some good points. Is it really as efficient as IBM suggests?
> What are the advantages that MVRC provide?

I love it when they mention stuff like:
"This does not conform to any ANSI standard isolation level, nor is it the way other RDBMS's work."

First: Their non-RDBMS databases used to go to great lengths to provide the isolation level that they "blame" Oracle for. That's how their Codasyl stuff works. And they go to great lengths to "prove" that in that "enterprise high-volume" environment it is absolutely essential. Sudenly with Oracle it's a "disadvantage"?

Second: Who the F*** cares that "other RDBMS's work different"??? Last I looked no one ANYWHERE required them to work the same way? Hellooooooo?

And "It is true that a reader will not wait on a writer but it is reading old, and possibly out of date data" is just the biggest pile of crap I've seen! Oracle reads the data as consistent with the start of the query as it can possibly get, guaranteed, and that is "possibly out of date data"?
Lights seem to be on, but is anybody home???

Conveniently, they "forget" to explain exactly how much more "efficient" their "Repeatable Read" is... What a joke!

The whole section on "Benchmark results" is laughable.

It looks like someone at IBM has woken up (FINALLY! It's only taken them what, 10 years?) to the differences in locking between DB2 and Oracle and is now desperately trying to cover up that their "database" needs to lock rows to ensure consistent reads!

Must be time for a new release of DB2, the hype is increasing...

-- 
Cheers
Nuno Souto
wizofoz2k_at_yahoo.com.au.nospam