Newsgroups: comp.object,comp.lang.c++,comp.lang.eiffel,comp.lang.java,comp.lang.objective-c,comp.lang.sather,comp.lang.smalltalk
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!godot.cc.duq.edu!newsgate.duke.edu!news.mathworks.com!uunet!in2.uu.net!ftpbox!mothost!schbbs!news
From: shang@corp.mot.com (David L. Shang)
Subject: Re: # ANNOUNCEMENT #: The Developer - A New Object-Oriented Language And IDE
Reply-To: shang@corp.mot.com
Organization: MOTOROLA 
Date: Tue, 16 Jul 1996 14:38:20 GMT
Message-ID: <1996Jul16.143820.25570@schbbs.mot.com>
References: <1996Jul16.133359.23152@schbbs.mot.com>
Sender: news@schbbs.mot.com (SCHBBS News Account)
Nntp-Posting-Host: 129.188.128.126
Lines: 37
Xref: glinda.oz.cs.cmu.edu comp.object:52094 comp.lang.c++:200391 comp.lang.eiffel:15250 comp.lang.java:68771 comp.lang.objective-c:5445 comp.lang.sather:2857 comp.lang.smalltalk:40862

In article <31EA8E43.1105@ix.netcom.com> Giuliano Carlini  
<giuliano@ix.netcom.com> writes:
> I agree with your sentiments regarding the desirability of GC, but
> must quibble with your statements. The hybrid nature of references
> doesn't generally affect whether a program can use GC. The exception to
> this is if the program has a huge heap; say 100Meg+. What does affect
> the decision are the external requirements placed on the program, and
> possibly the programming model.
> 
> For example, a multi-threaded program which must spend no more than
> 5% of it's time in the memory manager, and which must not pause longer
> than .05 seconds, and which must not allow garbage data to exceed
> 10% of the live data, and which runs on a CPU without virtual memory,
> probably can't use GC. It doesn't matter if pointers are hybrid or not.
>
> Another example: a program using a distributed persistent active object
> model may not be able to use GC. GC algorithms for such systems are
> being worked on, but this is still a research area. If you're bold you
> might use such a GC in a commercial project, but generally I'd suggest
> waiting. Once again, it doesn't matter if pointers are hybrid.
>

Probably you misunderstood my point. In fact, we have the same point of
view: GC is not the sufficient solution. The right solution is to use the
right reference, not a sigle reference protocol for all. The integration
of object references into the concept of object itself will enable such
complete solution. The distributed persistent active object model is
just one of many domain-specific examples that we want to cover. We cannot
stuff a language with various models, making it a super hodge-podge.
The solution is unification, or in other words, removing the hybrid
things. When the concept A and the concept B unified into a concept
AB, AB can cover a wider range than A+B. Reference unification is just
one of many unifications.

David Shang

 
