Newsgroups: comp.lang.smalltalk
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel-eecis!gatech!csulb.edu!news.sgi.com!howland.erols.net!vixen.cso.uiuc.edu!newsfeed.internetmci.com!in3.uu.net!news2.new-york.net!not-for-mail
From: vlad@world2u.com (Vlastimil Adamovsky)
Subject: Re: Another Way To Think About Smalltalk
X-Newsreader: Forte Free Agent 1.0.82
X-Nntp-Posting-User: (Unauthenticated)
Message-ID: <DyMn8r.7oM@news2.new-york.net>
References: <32507D8C.3B91@infinity.com>
X-Trace: 844223729/9879
X-Nntp-Posting-Host: i123.139.world2u.com
Date: Wed, 2 Oct 1996 02:35:38 GMT
Lines: 60

******* <123@infinity.com> wrote:

>After a lot of work with Smalltalk I've pretty much 
>concluded the present manifestations are worthless
>as application delivery systems. I see some hope from
>Smalltalk MT, but just a hope. 

I think too that Smalltalk MT goes in good direction. Since it is not
multiplatform Smalltalk, it can be "connected" to the underlying OS very tightly
without any wrapper classes designed for platform independence.
Thus, the performance is very good and still you can develop your applications
very quickly.

> but less well known is the fact
>that VS/Smalltalk Express (or any smalltalk with a good
>OS interface) can be used as a experimental test bed
>for studying OS features and exploring the "Internals"
>the OS has to offer. I was able to study Windows16 
>quite effectively in this way using what is now called
>SE. 

It is really a very good tool for exploring Windows OS. I have been using this
"tool" from the very beginning of existence of SmalltalkV/Win16.
It is great for exploring API's.

>Over the last two years I have often advocated deeper
>system (host) integration. 

It is correct, then we can compile direct to machine code, skipping VM. 
Is the binary portability really important? I don't think so. The source code
portability is much more important.

>The peculiarities and strengths of a machine have to be 
>emphasized, not glossed over. That is why the VM
>approach is doomed to failure. 

VM could be implemented as a "subsystem", that would be used when really
necessary.

>So the logical result of this argument--the opposite
>of VisualWorks or Java--is a Smalltalk based on
>a hardware processor specially designed for it. Indeed,
>a whole computer designed for Smalltalk, which would
>become the OS. A system built from scratch, 
>just for smalltalk. (OK, I would settle for a card.)

We do have Windows, which shares some simmilarity with Smalltalk.

>The Black Cube came very close to achieving this sort
>of integration, with Objective-C, but failed for 
>other reasons. 

The Black Cube came up ahead of time.



 Vlastimil Adamovsky
 ** C++ and Smalltalk consultant**
 **http://www.stepweb.com*********

