RE: Swarm -> Windows-NT...

[Steven Clark]

Consider Lisp.  Back in the good old days, people developed special
hardware to run Lisp programs.  However, around the mid '70s stock
hardware running less efficient but portable software Lisp environments
won out.  The large demand for stock hardware caused the time from
laboratory to product to be much shorter for stock hardware than for
Lisp hardware.  This resulted in the software Lisp environments running
faster because, although they were less efficient, they were available
on the latest generation of hardware, while the Lisp machines were
running last year's hardware generation.

The moral of the story:  if you can increase the user base of Swarm by a
factor of ten by sacrificing 10% of its efficiency, you will end up with
a net gain in efficiency.  The larger user base will drive a larger
Swarm development effort, which will give you a better Swarm than if you
applied less development effort to an initially more efficient Swarm.
(I realize I am blurring the difference between hardware development
effort and software development effort.  I think it is justified in this
case.)

Steven J. Clark       address@hidden       313-769-4396
Center for Electronic Commerce, Industrial Technology Institute
Box 1485, Ann Arbor, MI  48106


> -----Original Message-----
> From: glen e. p. ropella [SMTP:address@hidden
> Sent: Saturday, May 17, 1997 7:48 PM
> To:   address@hidden
> Subject:      Re: Swarm -> Windows-NT...
> 
> [Steven Clark]  [...] 
> 
> We may take a performance hit.  But, one of the points that 
> de-emphasizes the perf. hit is that people with large and 
> compute intensive simulations can run in batch mode and play
> back any data they may need in whatever viewer they can.  That
> doesn't eviscerate the perf. hit point because it means that 
> Swarm wouldn't be ideal for high interactivity and high performance.
> 
> But, I think the gains we will make in the broadening of our 
> user community to include the WinNT world *and* the Java
> world outweighs the perf. hit.  And, hopefully, using Java
> for our GUI will open up new possibilities for remote execution
> and performance optimizations later on.
> 
> Any flaws in my reasoning?
> 
> glen
> 

                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.
                  ==================================