Re: Changing isolation level?

NoJunkMailshah_at_xnet.com wrote:
>
> First transaction reads a row and deletes it, the changes has not yet
> been commited.

This sounds like something Advanced Queuing may be suited for.

> Meanwhile the second transaction tries to read a row and get same row
> again, it has not yet deleted the row.
>
> First transaction commits the changes.
>
> When second transaction tries to delete the row it gets error.

Fine. It gets an error: it's not like your server catches on fire. Catch the exception and deal with it.

>
> This is a part of big application and has worked with VSAM and DB2 where
> they lock the rows once someone updates or deletes it but has not yet
> commited.

Do VSAM and DB2 block when they get to the locked row, or do they just skip it?

>
> I may be wrong here, but if the row has been (marked for) updated or
> deleted and another transaction get old value then I consider it a dirty
> read.

I dirty read is reading uncommitted data. It aint' doing that.

> This may be O.K. in data warehouse but is not acceptable in online
> transaction.

> Imagine in a banking environment, you are withdrawing money from an
> account at the same time your spouse at another ATM is also trying to
> withdraw the money and he/she gets wrong balance.

OK, I'm imagining. And in fact, I'm pretty sure that banks don't guarantee that the balance given reflects even all recent transactions, much less all uncompleted transactions, so I'm betting banks use an Oracle-like model.

Xho

-- 
-------------------- http://NewsReader.Com/ --------------------
Usenet Newsgroup Service              New Rate! $9.95/Month 50GB