[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: A path for CSP-based Solutions towards HUGE Industrial Success.

Larry Dickson wrote:

> >  o CSP is remembered from college days - if at all - as some pretty
> >    hard maths ...
> But those simple diagrams you did of hardware-equivalent components
> are not so hard.

Right - of course they are not hard.  But those diagrams of hardware-like
components are not the way CSP has been presented at most universities.
You get that if you were a student at one of the WoTUG community colleges,
but elsewhere you get the (hard) maths and, despite Dijkstra's best efforts,
most systems designers and programmers are not mathematicians :-).

The beauty of occam (and, with care, the CSP libraries for C, Java etc.)
is that the designer/programmer works with the CSP model and gets all the
benefits from its clean compositional semantics *simply by using it*.  It
untangles the mess that otherwise exists between serial and multithreaded
coding and that's an enormous win.  But very few have experienced this
(since the demise of the transputer) ...

BTW Marcel: I agree with you that CSP provides (possibly) its biggest win
in the area of concurrency within a single processor.  When I said "transputer
and associated hardware/software technologies", I did not mean to imply
physically parallel computing - the transputer is an idea, not a chip!

Mind you, CSP will provide some really big wins for hardware design and for
hardware/software co-design.

And the *lightness* (e.g. 100 nano-second context switch) with which the
CSP primitives can be managed also offers big wins for distributed computing
(e.g. scalable servers) ...

Now, again, who wants to be a millionaire ...