Path: utzoo!utgpu!water!watmath!clyde!rutgers!sri-unix!husc6!bbn!gatech!ukma!psuvm.bitnet!uh2
From: UH2@PSUVM.BITNET (Lee Sailer)
Newsgroups: comp.databases
Subject: Re: Data vs Application oriented design
Message-ID: <29538UH2@PSUVM>
Date: 11 Jan 88 21:31:17 GMT
References: <12673@pyramid.pyramid.com> <69@coot.AUSTIN.LOCKHEED.COM>
Organization: Penn Sate Erie--School of Business
Lines: 34

In article <69@coot.AUSTIN.LOCKHEED.COM>, chris@AUSTIN.LOCKHEED.COM (Chris Wood) says:
>
>Thus it follows that the Relational model is a SUBSET of the network model.
     
I can sort of agree with that.  It is similar to saying that Modula 2 is
a subset of Ada.  One language philosophy is to build in EVERY feature
that might ever be useful.  A different philosophy is to build in
JUST ENOUGH features to make everything possible.
     
The Relational model provides a subset of the features, but still allows
one to build whatever you need to build.
     
Why is this minimalist approach popular?  Well, it is a lot easier to
implement a minimal system and be reasonably sure that it is correct.
Look how much easier it is to create a M2 compiler than an Ada compiler,
and forget about COBOL and FORTRAN.  Likewise, a new computer system will
see its first Relational DBMS long before its first Network system.
     
     
BUT!
     
The *ideal* relational system also supports some capabilties that the Network
model does not, such as seamless integration of database and data dictionary.
Also, much of the modern query optimization depends on the use of a
minimal, mathematically coherent set of operators.  As soon as
ad hoc (though "useful") features are added to the language, optimaization
techniques begin to fail.
     
Languages are designed to be easy to use.  They are also designed to be
easy to port, and to produce efficient code.  Some ease of use is worth
sacrificing if it increases portability and efficiency enough.
     
Remember, too.  Anything currently available is already obsolete.