Hi Gang,
    Respectfully, I must disagree. The idea of modular software goes
back to the idea of "Software ICs". If the hardware guys can glue chips
to boards and backplanes in order to build up systems, why can not the
software guys plug modules into a core environment in a similar
fashion? Indeed! It sounds so simple!
    Clemens Szyperski, one of Niklaus Wirth's grad students, wrote a
book a couple of years ago: Component Software, Beyond Object-Oriented
Programming, Addison-Wesley, 1997-1998, ISBN 0-201-17888-5. Szyperski
was a co-founder and Director of Research at Oberon Microsystems, Inc.,
which eveloped the BlackBox Component Builder, one of the first development
environments designed specifically for component-oriented programming.
Szyperski also developed Component Pascal, an Oberon derivitive, while
at OMI. Szyperski is now a member of Microsoft R&D.
   The first part of Szyperski's book goes into the gory details of
components (software modules) versus object technology. I can't even
begin to discuss the issues therein at this point. The second part
of his book is a deep discussion of OMG (COORBA and OMA) vs Microsoft
(DCOM, OLE, and ActiveX) vs Sun (Java and JavBeans) and ends with
a number of open problems. The final two parts of this book deal
with next gnerations and marketing considerations.
    Interestingly enought, Szyperski cites Apple's OpenDoc and his own
BlackBox Component Framework as somewhat similar and claimes that the
two systems are two of the few that were ever developed to product level.
The BlackBox Framework was based upon Hierarchical Model-View-controller
(HMVC). Of course, OpenDoc failed and was discontinued by Apple in 1997.
The supporting Component Integration Labs plus the Live Object
certification and validation services were also terminated at that time.
For the really adventuresome, the OpenDoc software may still be available
from IBM.
    Unfortunately, Szyperski learned his lessons well from N. Wirth:
I had to read his book a couple of times before I could even begin
to understand the issues involved. But I did gain an understanding of
why one just can't easily plug modules/components together and
into some suitable core system in order to build up a viable system,
particularly, one that is object-oriented. So, I stick to my original
assertion: modular software or component frameworks are a swamp that
exponentially deepens and broadens beyond one's imagination. Don't go
there without Big Bucks and Cubic Time!
   To end on a more positive note, Florin X Mateoc had a great idea
which I want to comment upon in one more post and then I stop. :-)
Cheers, Roger.....

Dirk Wessels wrote:
> 
> ----- Original Message -----
> From: Roger Vossler <rvossler at qwest.net>
> To: Squeak List <squeak at cs.uiuc.edu>
> Sent: Saturday, March 10, 2001 7:59 PM
> Subject: Modular Squeak
> 
> > Hi Gang,
> >
> > "Large monolithic images" versus "a small core plus a collection of
> > modules or plug-ins" are two very conflicting design approaches.
> 
> They are not conflicting.
> 
> I already had proposed my "views" but I see that you guys and girls have not
> understood it well enough. Or I have not explained it clearly enough...

[snip]