Re: does a table always need a PK?

From: Paul Vernon <paul.vernon_at_ukk.ibmm.comm>
Date: Mon, 1 Sep 2003 22:11:54 +0100
Message-ID: <bj0d5h$1fbc$2@gazette.almaden.ibm.com>

"Tony Douglas" <tonyisyourpal_at_netscape.net> wrote in message news:bcb8c360.0309010259.41a03bb3_at_posting.google.com...
> Hi Heikki,
>
> "Heikki Tuuri" <Heikki.Tuuri_at_innodb.com> wrote in message
news:<gZG3b.333$st4.19_at_read3.inet.fi>...
> > in computer science people usually think that a formal definition has to be
> > mechanically verifiable, at least in principle.
> >
> Indeed, and where possible this should be the approach taken. The big
> challenge is, how do you get mathematics to say something in any way
> useful about things which are partially policy (backup) and partially
> down to greasy bits of machinery (tape drives) ?

I'm sure that some math/thoery could be/has been built that takes into account parameters such as MTBF and other 'risk factors', combined with requested security margins (1 in X chance of loosing 1 bit of data) so that some rigour can be added to stuff like backup policies.

Regards
Paul Vernon
Business Intelligence, IBM Global Services Received on Mon Sep 01 2003 - 16:11:54 CDT