Re: Concurrency Questions

From: David Cressey <david_at_dcressey.com>
Date: Wed, 17 Jul 2002 12:10:19 GMT
Message-ID: <L8dZ8.2$Dt2.329_at_petpeeve.ziplink.net>


CN,

> I don't want to write programs which can cause abortions because I
> think it is complicate to clean up things when abortions arise.

A good applications designer, working with a good RDBMS, will always have an effective strategy for
preventing harmful concurrent access to data, with appropriate ways to cause the transactions to be performed
serially, or to produce a deadlock in the worst case.

With the best of the commercial RDBMS packages, in the scenario you outline, the applications programmer will
have to plan for deadlock, and have code to recover from deadlock when it occurs.

It's not clear to me whether your original question was about how to design the application, about how to design the RDBMS package in the first place, or about how to get the right result at the applications level when the RDBMS package provides inadequate support for concurrency control. This may be clear to other readers. If it's not clear, then you will get better answers if you clarify.

Good Luck,

Dave Received on Wed Jul 17 2002 - 14:10:19 CEST

Original text of this message