Re: computational model of transactions

From: Cimode <cimode_at_hotmail.com>
Date: 1 Aug 2006 15:00:16 -0700
Message-ID: <1154469616.293989.27990_at_i3g2000cwc.googlegroups.com>

paul c wrote:
> Cimode wrote:
> > paul c wrote:
> >> Bob Badour wrote:
> >>> paul c wrote:
> >>>
> >>>> Bob Badour wrote:
> >>>>
> >>>>> paul c wrote:
> >>>>> ...
> ...
> >> Like give all 100,000 employees a 10% raise. Still, that kind of commit
> >> is not what I call a logical commit, suggesting that a commit doesn't
> >> mark a luw boundary. I've heard it called an 'intermediate', aka
> >> physical, commit.
> > I understand and agree with this last point. As I mentionned this is a
> > complex problem and a simplyistic solution just won't do it.
> > ...

>

> Sometimes it is actually handled, at a sort-of application level, by
> adding another attribute, say "update not finished" which all the rest
> of the transactions must respect, so as to prevent somebody from getting
> a 21% (1.10 X 1.10) raise if the power fails. Or a 'shadow' table which
> is later quickly swapped via some physical indirection. As far as RT is
> concerned, time is physical, not logical, no matter what anybody else
> says, ha ha!
I agree. I just don't feel confortable with the concept *handled* on a logical perspective. RM has some logical tools to deal with the question but they seem unsufficient.
> p
Received on Wed Aug 02 2006 - 00:00:16 CEST