Newsgroups: comp.lang.smalltalk,comp.object
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!news.alpha.net!uwm.edu!math.ohio-state.edu!howland.reston.ans.net!pipex!warwick!bham!bhamcs!comlab.ox.ac.uk!sable.ox.ac.uk!lady0065
From: lady0065@sable.ox.ac.uk (David Hopwood)
Subject: Re: IDL Lobotomy (was: CORBA...SmallTalk...)
Message-ID: <1995Mar14.171752.19206@inca.comlab.ox.ac.uk>
Followup-To: comp.object
Sender: david.hopwood@lmh.ox.ac.uk
Organization: Oxford University, England
References: <CLINE.95Feb11195907@sun.clarkson.edu> <3k0e4a$ob9@picasso.cssc-syd.tansu.com.au> <3k1pmd$e23@bcarh8ab.bnr.ca> <3k44ae$aev@news.irisa.fr>
Date: Tue, 14 Mar 95 17:17:52 GMT
Lines: 50
Xref: glinda.oz.cs.cmu.edu comp.lang.smalltalk:21739 comp.object:28084

[comp.lang.smalltalk removed from follow-ups]

In article <3k44ae$aev@news.irisa.fr>,
Jean-Marc Jezequel <jezequel@stormbringer.irisa.fr> wrote:
>|> >edwards@world.std.com (Jonathan Edwards) wrote:
>|> >>
>|> >> What I wonder about HP-DST is why anyone would want to lobotomize
>|> >> Smalltalk down to the level of C++ by imposing IDL on it. 
>|> >> --

I don't know HP-DST, so I won't comment on this.

>Just wondering if the OMG/IDL/CORBA people have ever heard of the problem of
>the Inheritance Anomaly occuring in Object-Oriented Concurrent Programming
>Languages (1). Since there are no *referred* scientific publication on CORBA,
>it is hard to judge.
>But from my limited knowledge of public CORBA documents, this problem is not
>tackled seriously.

No, as far as I'm aware it isn't even mentioned. But then the aim of CORBA
is simply to allow message passing between objects defined in different
environments; not to allow inheritance between them. So the inheritance
anomaly only arises within the particular languages that are used to
implement objects.

This isn't necessarily a serious problem. The flexibility of object based
systems without inheritance (but with dynamic binding and polymorphism)
is often underestimated - possibly because there are very few of these systems.

>This implies that IDL are useless, and even harmful, to
>deal with OO distributed systems. On the other hand, if you're not interested
>in inheritance (e.g. you use C++--) it makes sense to use it.
>
>Conclusion => CORBA/IDL is not object oriented, just object based.

So it's object based. Doesn't mean it's not useful.
(BTW, C is not an object-based language).

>And this is not to count with the problems Greg Wilkins
><gregw@ind.tansu.com.au> wrote about:
>
>>For heavily-used polymorphic methods, this transformation causes
>>an EXPLOSION of interfaces - which in turn means that managing the
>>'name space' of one's interfaces is much more difficult.

This can be a problem. What IDL needs, I think, is signature typing, and
possibly stub interpreters rather than compilers.

David Hopwood
david.hopwood@lmh.ox.ac.uk
