On Thursday 27 July 2006 12:18, David Binger wrote:
> On Jul 26, 2006, at 9:01 PM, David Binger wrote:
> >> In practice, I am only interested in what someone else is doing if
> >> I have
> >> already read the data and I now want to commit a change.
> >
> > That seems reasonable.
>
> Looking at this today, I think I should add that
> MVCC, as I understand it, will provide what you
> are asking for here, but that may not help the application
> you are describing if it continues to require rapid
> writes to the same instance.  Those rapid writes may cause
> conflicts anyway.  With MVCC, you avoid the immediate ReadConflict,
> but you will get a conflict later if you try to commit anything
> before you call abort.

I agree that MVCC will do the job. Depending on how you propose to implement
it I have a suspicion that it might be more than I am asking for, but I need
to sit down with a wet towel round my head before I can come back on that.

As for the rapid access leading to conflict anyway, I also agree, but these
would always be genuine conflicts instead of the often fake conflicts that I
believe I see now. However, the solution here is application design and I
have that in hand. I do wonder if Quixote needs to update the access time
every time, but that is a question for another list :-)

Thanks for the help and comments.

Mike S.
--
Mike Sandford
Director
Campbell Carr
12 Broadway
Amersham
HP7 0HP

Tel: +44 (0)1494-432323
Fax: +44 (0)1494-434888
Mob: +44 (0)7946-710026

www.campbellcarr.co.uk
m_sandford@campbellcarr.co.uk