Re: Divorcing data storage from business logic

[Paul Makepeace]

On Tue, Jul 03, 2001 at 11:57:53PM +0100, Matthew Byng-Maddick wrote:
> Erm DB3 does transaction locks. All of the DBM routines have used locking
> and hence concurrency has not been an issue.

The issue isn't whether you have it or not but whether they render the
thing useless when 'n' instances of a program are fighting over the
lock.

> This is more than can be said for MySQL with the horror that is MyISAM.

I don't use MySQL :-)

> > I just wouldn't bother. Go straight to the RDBMS. Bear in mind in some
> > cases (notably Oracle, and MySQL can be made to fly too) Perl is the
> > limiting factor, not the database. YMMV, IMO, etc
> 
> :-) This can be the case. It depends on your disks.

Not really, simply moving the stuff around in a Perl hash is slow.
Thanks to caching (data, queries, indexes, execution plans, etc) disks
don't really come into it unless you're doing some pathological search
or you're contending with lots of DSS users hammering it to death.
(Unless you mean a x686 using an 5MB ST drive :)

Paul

-- 
A very small object
-Its centre