Re: Dreaming About Redesigning SQL

"Anthony W. Youngman" <[EMAIL PROTECTED]> wrote in message news:<[EMAIL PROTECTED]>...
> In article <[EMAIL PROTECTED]>, Lauri Pietarinen
> <[EMAIL PROTECTED]> writes
> >>The "Little Languages" approach has tended to be more successful, as
> >>it doesn't force everyone into one guaranteed-to-be-wrong-sized
> >>Procrustean bed.
> >>
> >This is an interesting subject.  Now when I think more of it,  it really depends 
> >on how complete this
> >unified environment would be.  Is it possible to have a simple language that 
> >does nearly everything
> >you need?
> 
> They've tried. And ended up with abhominations like PL/1 (which is
> actually not a bad language at all - it just suffers from trying to be
> all things to all men, and it shows! :-(

Yes, I know.  I used it for about 12 years.  It's a good language if
you stick to a strict subset and don't get lost in all the esoteric 
features.  It had (more or less) decent error handling already
in the 70's!

> 
> I would suggest trying to become an expert in languages from different
> FAMILIES. Then you'll see why it pays to be master of many.

What I sense is that people are getting very frustrated with having to
use so many different kinds of environments to get ONE application
running.  It might not be a problem to somebody clever like you, but
for large organisations buiding large systems it is becomming a
nightmare.

Anyway,  PL/I did not address the database side, you had to use DL/I
or embedded SQL.

So my guess why many programmers dislike SQL (and hence, databases) is
that they have to deal with two different paradigms.  We end up with
J2EE, where everyting becomes a bean of some kind and we hide simple
SQL in the beans without leveraging the power of the DB.  Guess what -
we get performance problems!!

Or we end up with programmers building their own DBMS'es so as to at
least get an illusion of unification.

Now, the alternative COULD be to widen the domain of the DMBS.  It has
lot's of information that could be used to create or generate (even
dynamically) lot's of the application, i.e. table definitions, columns
and domains.  By adding metadata we could even generate titles for
fields and screens in user interfaces.  Using constraints, most of the
checking can be done by the DMBS and it would be automatically applied
to all applications. This is an area that has been hardly scratched.

regards,
Lauri Pietarinen