Re: Quote from comp.object

On Mar 5, 7:48 pm, "DBMS_Plumber" <paul_geoffrey_br..._at_yahoo.com> wrote:
> I could be convinced of this. It's an attractive idea, but speaking
> purely as a plumber, you're going to have a very hard time convincing
> me that higher level 'functional' languages are a practical platform
> for implementing something like a DBMS. DBMS internals do all kinds of
> evil, ugly things purely for performance.

I doubt that we'll ever see a DBMS actually implemented in a functional language, although it would be very reassuring to see that it could be done as a theoretical exercise, and for proving correctness of fully optimised production DBMSs, which as you say have to pull all sorts of tricks to wring as much perfomance as possible from the available hardware. It's the old trade off between the abstraction level you want to work at, versus the need to work at the level that gets most performance.

> It ought not be this way --
> Hell! My first love was LISP --

You have my sympathy ;) I'd do everything in Prolog if I could get away with it, and it had a type system worth a candle.

> but each time someone has made a foray
> into systems programming using higher level languages we rediscover
> the virtues of lower level languages. It has nothing to do with the
> theory. It's that higher level languages place more demands on the
> quality of the compiler / interpreter.
>

Although to be fair, they are getting better and better all the time.

> Also, my limited experience suggests that the overhead of coordinating
> the 'hundreds of thousands of simultaneous processes' begins, at some
> point, to swamp the marginal return of adding another concurrent sub-
> task.
>
> Of course, I still program in 'C' using 'vi'.

'vi'. Now there's *my* idea of an IDE !

• Tony