Re: does a table always need a PK?
Date: 28 Aug 2003 04:07:21 -0700
Message-ID: <bcb8c360.0308280307.7293d31a_at_posting.google.com>
Hello there,
"Heikki Tuuri" <Heikki.Tuuri_at_innodb.com> wrote in message news:<Xr63b.703$n62.374_at_read3.inet.fi>...
> yes, you understood what I meant.
Good-o.
> It is rather easy to formalize that the
> query language must be able to perform the operations of the relational
> algebra, since the relational algebra is a simple mathematical system.
So the immediate question that springs to mind is, so why did DBMS
vendors do SQL instead ?
> But a DBMS is a kind of operating system, too. And it is hard to define
> formally what required concurrency, backup, sysadmin, communications, etc.
> handling capabilities are for a program to be called a DBMS.
>
Indeed. Which is why I had issues with the initial answer, requesting
something which I suspected we both knew didn't currently exist and
would be non-trivial to formulate. I would ask, though, why we would
be asking for a mathematical def'n of something about which
mathematics currently has little or nothing to say ? (Or has someone
produced a mathematical model of a tape drive, how to write to it, and
what a backup to it means, for example ?) Or put more simply, do all
formal def'ns have to be mathematical def'ns, or are there some
problems for which we have to descend to natural language (with all
that entails) whether we like it or not ?
> About the horse: if we genetically engineer a horse and add DNA from a frog,
> at what point does the creature transform from a horse to a froghorse?
>
An interesting question - which leads to other interesting questions
about how pure horse DNA has to be at any given time to qualify as a
horse (e.g. are only thoroughbred racehorses with provable ancestry
really horses ?) Not being a geneticist I'm not going there - I know I
don't know enough about that to hold a reasonable discussion on it !
- Tony