[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: The world needs process-orientation
Hi Neil, Denis, and all,
This refers also to Ruth Ivimey-Cook's thread "Re: CPA 2006 - Call for
Papers" and my thread "Re: Is OO a deliberate fraud?".
Neil, you have identified what ought to be a great value proposition for
> Interesting times. Just as I find one mailing list discussing the
> future of process-orientation (my preferred umbrella term for CSP and
> similar ideas), I find another crying out for it. Game Development
> Algorithms is a list I was pointed to some time ago when someone
> (Richard, I forget his surname!) had pointed the list at my C++CSP library
> as an example of a different approach to concurrency (i.e.
> process-orientation). At the time I came close to veering off-topic and
> the discussion petered out. Fast-forward a few months and another
> discussion ends up on the topic of concurrency again, and the usual
> problems are bandied about - threads too heavyweight, too dangerous, too
> hard to reason about - but with no good solutions being suggested.
The problem is IT DOES NOT MATTER IF WE OFFER SOLUTIONS WORTH BILLIONS OF
DOLLARS. WE WILL NOT BE LISTENED TO. We are in a positive-feedback death
spiral, created by the OO people who commandeer all the relevant
terminology, apply it metaphorically to unworkable irrelevancies, and use
their clout to enforce their interpretation. Look at Ruth's thread:
"Denis A Nicole" <dan@xxxxxxxxxxxxxxx> wrote:
> On Tue, 23 May 2006, Ruth Ivimey-Cook wrote:
>> We must put our best foot forward on this: the advertising so far (I'm
>> aware of Peter's messages on occam-com, java-threads) and my postings to
>> various internet newsgroups - c.s.transputer, c.parallel, c.realtime
>> among them) have resulted in a mere handful of different host addresses
>> accessing the website. Even if all of those came we'd be in dire
> I'd really like to come. I could probably even find something interesting
> to present, but I simply cannot justify the £475 registration fee. As we
> are now such a minority, I don't believe we can accommodate the standard
> University gouging conference rates. It's all very well for GGF, with
> eScience money to burn (although I've complained there too), but a
> pressure group like us can't afford it.
And he goes on to suggest camping! Exact historical analogy: the epicycle
people have WON, and Kepler must camp out in the brush if he wishes to
address a group of scientists...
> Needless to say I couldn't resist discussing process-orientation again.
> Games are an interesting area; they have historically been concerned about
> performance above all else. This has led to the recent consoles having
> major parallelism inside - the Xbox 360 has 3 CPUs, and the PS3 has the
> Cell. So games developers will find themselves forced to use concurrency,
> but are discovering the usual problems with it. The
> Sweeney talk (linked to below in my other email) is very interesting, I
> suggest you all read it if you are worried that what we do is becoming
> less relevant rather than more; the latter is very much the case.
I have pursued the game development / highly parallel CPU opportunity for
two years, including a white paper, written this year, showing how to deal
with their problems. When I was interviewed for a job by a major game
console manufacturer, the technical people were very interested, but I was
vetoed by top management, who are married to their non-working design.
Quality is actually a threat.
We are locked out of commercial and government support by the investors'
fear to move beyond their OO-based experts. There used to be
off-the-beaten-track investors, so-called entrepreneurs who would place
money based on their good judgement, but these seem to have died out (I
just got a response from one, declining to talk because it was "over his
head"). We need to get real about saving our work. Should we head for
campsites and do without investment? Pool our funds? Everybody, please
come up with ideas! As Bilbo Baggins would say, this is what our council
has to decide, and all it has to decide.
Here is my 2 cents worth: It needs to be a territorial thing of
process-oriented people. We have to fence ourselves off from "standard
computer science" or the OO hordes will dilute our directions before they
can reach critical mass. This is the sort of role Inmos used to play. Our
territory must offer hardware, OSs, compilers, linkers - all
process-oriented only (very powerful hardware, by the way, now costs a
hundred times less to develop than it used to). Only in this way can we
make our case against the will of the managerial empires.
> While I would love for them all to start using occam or
> the like, perhaps (excusing my self-indulgence) a port of C++CSP to the
> Cell and 360 is more the approach that is needed in the short-term.
Writing an occam compiler for the Cell or 360 ought to be within even our
diminished capabilities. Has anyone done it?