Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!rutgers!uwvax!tank!mimsy!eneevax!haven!ncifcrf!nlm-mcs!adm!xadmx!joel@wsl.dec.com
From: joel@wsl.dec.com
Newsgroups: comp.lang.pascal
Subject: Re: Pascal dying out?
Message-ID: <17360@adm.BRL.MIL>
Date: 26 Oct 88 02:50:51 GMT
Sender: news@adm.BRL.MIL
Lines: 31


A. R. Thompson has a few misconceptions about Modula-2.  The language certainly
has some problems, but none that he describes.

First, the type WORD is imported from a machine-dependent module SYSTEM.  Using
WORD is therefore not portable to all other implementations, but is the right
thing for the machine at hand.  In particular, on a VAX the type WORD is 32
bits long.

I have been convinced that structural compatibility of types is not bad in all
cases, particularly if you are dealing with data structures residing on remote
machines.  In this case it is hard to maintain name compatibility.  But in any
case, Modula-2 uses name compatibility.

Modula-2 allows the declaration of modules inside of procedures, but hardly
anyone uses them this way.  Most modules are declared as compilation units, and
data declared in these modules exists for the life of the module.  In most
systems, this is synonymous with the life of the program, but in systems that
support dynamic loading the module data may live longer.

Modula-2 modules allow the definition of ``abstract data types'' in the original
sense of the word.  It does not support subtyping or object-oriented programming
directly, though many have kludged around this by embedding procedure fields
inside of records.  You do give up type-safety if you want to subclass using
this technique.

Modula-2 is a much nicer language than Pascal, with none of the problems
attributed to it by A. R. Thompsen.  I can't speak for the extended Pascal
proposal.

- Joel McCormack