Newsgroups: comp.sys.transputer
From: NewsMan <2021@please.dont.send.email>
Subject: Re: Post Mortem
Date: Sun, 22 Feb 1998 06:10:11 +0000
Mime-Version: 1.0
Message-ID: <q1n8mGADF870Uw3C@spyglass.demon.co.uk>

David Boreham <dboreham@netscape.com> wrote
>The point about the transputer being someone's science
>project is well made. Mind you, if the science project had
>worked, solving the problem of how to program a distributed
>multiprocessor machine, the payoff would be huge.
>Every day I wish I knew how to solve that one.

Well, the answer from the original designers seemed to be CSP. Or do you
mean something else? Sticking my neck out a bit here, to my mind we
could do with a process-farming system (to keep as much of the
computing power as possible constantly busy), dynamic allocation of
processors to tasks by the loader, protected and relocatable (swappable)
memory regions, and objects defined by writable data regions with
reentrant code. This is partly down to operating system design but the
above could run any well known o/s with the distributed nature of the
system invisible to the user. Comments?

NewsMan