Re: optimal size for rollback

Nuno Souto <nsouto_at_optushome.com.au.nospam> wrote in message news:<3d8dc82d$0$22173$afc38c87_at_news.optusnet.com.au>...
> 22 Sep 2002 04:23:35 -0700, Igor Laletin said (and I quote):
>
> PMFJI
>
> >
> > I absolutely agree with you. Just such applications are rare. They may
> > exist in a training course but don't survive in production environment
> > for long.
>
> I'm going away on holidays soon, so I can't go into this like it
> deserves. Let me just say this: after using Oracle since 1986, I have
> yet to see out there a "properly designed" application that doesn't
> exhibit one or more of the symptoms Howard and Karen talk about.

There is no "more". We're talking about just one particular symptom - an application allows user to start a transaction and go for a lunch.

So most of the application you worked with since 1986 do that? Bad luck.  

> This leads me to think that the mythical "perfect application" just
> doesn't exist. What does indeed exist is a mish-mash of code that is put
> together by people who don't have a clue and in most cases in a way that
> is purported to be "portable" and is nothing of the such. The result?
> Applications that regularly break all the "rules" of good database use.
> And like it or not, we have to live with them.
 

> Now, OPTIMAL may be a great idea for a DBA that doesn't want to get too
> involved. Yet, it causes problems. Well documented. 1555 is one of
> them. The other is considerable overhead and disruption of smooth
> processing. For development databases it may make sense. For
> production, no way. Regardless of the application.

I can't say it better than Richard did. Just read his post not from the middle (disk prices - and yeah, they're expensive) but from the beginning. Received on Mon Sep 23 2002 - 02:10:19 CDT