Newsgroups: comp.lang.misc,comp.programming,comp.software-eng,comp.ai
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!mvb.saic.com!eskimo!usenet
From: hgeorge@eskimo.com (Harry George)
Subject: Re: A new approach to software engineering!
X-Nntp-Posting-Host: tia1.eskimo.com
Message-ID: <D21EF7.BF6@eskimo.com>
Sender: usenet@eskimo.com (News User Id)
Organization: none so far
X-Newsreader: WinVN 0.93.10
References: <D0wsx8.6zC@prl.philips.nl> <johnson.788537339@hal.cs.uiuc.edu> <3eb0je$dje@jhereg.mdd.comm.mot.com>
Mime-Version: 1.0
Distribution: na
Date: Sat, 7 Jan 1995 13:21:06 GMT
Lines: 66
Xref: glinda.oz.cs.cmu.edu comp.lang.misc:19856 comp.programming:13819 comp.software-eng:29283 comp.ai:26300

In article <3eb0je$dje@jhereg.mdd.comm.mot.com>, 
dcampbel@jhereg.mdd.comm.mot.com says...
>
[deleted material]
>Just like to add in my two bits on this.  In  my experience with 
>systems implementation and testing, the most common problem groups are
>those associated with the use of non-turing languages.  They are 
>always problematic in that user expectations often exceed the system's
>capabilities.  Expert systems, written in DbaseIII are doomed to fail,
>and expensively too.  Use of non-turing languages also guarantees 
>long-term system inter-operability problems: semantic translations
>have to be performed, etc.  
>
>I also would contend that a lack of such completeness was the basic 
>cause for the less than overwhelming success of CASE technology.  
>

I agree 100%.  Several years ago I was managing a data modeling and
process modeling organization.  We used all the best CASE tools.
The more we tried to accurately (no, let's say _adequately_) model
the clients' domains, the worse the tools looked.  I started looking
for alternatives and realized that a good, readable OOPL
was semantically more powerful and syntactically more elegant for 
exactly the same job -- and could be type-checked and executed as well.

After spending a very frustrating year trying to convince others of ths,
I gave up and changed organizations.  The CASE mavens are still at it,
with no more success than before.

>I have used both prolog and scheme in conjunction with c/asm to 
construct 
>numerous test simulators and other widgets, and have found that a 
>mixture of turing-complete languages provides me with the strongest
>compromise of interoperable tools with differing levels of abstraction.
>
I agree turing-complete is necessary, but I don't think it is sufficient.
Modula-3's rendition of objects, branded types, and partial revelation
are useful syntactic sugar which make the model more tractable.
Add to that hypertext links, and it begins to be useable.  I haven't
found any better combination of elegance and power.  

>I realise that this is a complex solution, but I am not a maker of 
>languages. 

Niether am I.  But I sure am glad there are languages which provide:
     both static and dynamic typing (with cross module checking)
     multi-paradigms:C-like bit twiddling, ADT's, Objects, 
         Lisp-like polymorphic lists and lambda creation
     modularity, with partial revelation and branding
     garbage collection
     parameterized exceptions
As I say, right now my favorite is Modula-3.

>
>: Ralph Johnson -- johnson@cs.uiuc.edu
>

-- 
Harry George
email: hgeorge@eskimo.com
smail: 22608 90th Ave W / Edmonds WA 98026
quote: "The reasonable man adapts himself to the world;
       the unreasonable one persists in trying to adapt
       the world to himself.  Therefore, progress depends 
       on the unreasonable man." G. B. Shaw

