Re: "Transactions are bad, real bad" - discuss

From: Paul Vernon <paul.vernon_at_ukk.ibmm.comm>
Date: Fri, 9 May 2003 11:02:52 +0100
Message-ID: <b9fug7$1s80$1_at_gazette.almaden.ibm.com>


"Mikito Harakiri" <mikharakiri_at_ywho.com> wrote in message news:xIUta.16$Im6.169_at_news.oracle.com...
> I'm not sure what is the best point to rollback:-)
>
> Did you agree that multiversion concurrency model and derived "flashback"
> query functionality allows quering the state of the auction AS OF 12:00 when
> physically queried at later time, say 12:15? It might seem to be no big
> deal, because alternatively the auction organizers must schedule their query
> exactly at 12:00.

> However, as you mentioned, if auction bidder is allowed
> direct access to the database,

Which is required for strong application independence.

> he might start transaction that
> [intentionally] would not meet the deadline.

Agreed.

> At 12:00, with one transaction
> still in progress organizers query would wait for an exclusive lock.

Yes

> You
> wont be able either to enforce the rule "Any transaction must be committed
> before 12:00", nor get the auction snapshot at 12:00.

Indeed. The business rule "the auction must close at 12:00" is not enforceable as a relational constraint is transactions are allowed.

Therefore if you want strong application independence, you cannot also have transactions.

Have I not proved my point?

Regards
Paul Vernon
Business Intelligence, IBM Global Services Received on Fri May 09 2003 - 12:02:52 CEST

Original text of this message