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

RE: It was a stressing weekend!


Lea' book discusses patterns, but many patterns are not really design
patterns. Instead, I read them as techniques or implementation patterns for
Java. Lea's book does not always make clear when patterns should be used in
design. However, it is a good book but you do not learn to design (and
implement) reliable software.
We learn from occam to design our programs in terms of communication and
compositions of processes. Many people think that communication and
compositions of code (threads) can be done in many ways, so, why should we
learn occam? I think that those people think that CSP is married with occam.
No, JCSP and CTJ prove that the CSP concept can marry with any
"object-oriented" programming language, such as Java. CSP is a way of
thinking, a way of designing, and a way of implementing. Also, the CSP
concept does not conflict with traditional programming. It is not one or the
other. The Java community should learn about compositional programming --
i.e. learning the CSP concept. For example, we need a book with the title
"Compositional Programming" and subtitle "Reliable and real-time software in
Java". The term CSP should be mentioned at the introduction and not in the
title. This book should teach us when and how to use patterns of channels
and processes. This book should be published by Sun Press next to Lea's

The NCIST organisation and Sun are developing Real-Time Java. But JCSP/CTJ
is Real-Time Java. They don't see the benefits of JCSP/CTJ (i.e. the CSP
concept) because they don't get enough information from us. They are
reinventing the real-time wheel for Java. I belief that Real-Time Java by
NCIST and Sun will result in an operating system API for skilled programmers
and not for the ordinary Java programmer (such as control engineers).
JCSP/CTJ is created for the ordinary Java programmer. Can we participate in
Real-Time Java?


-----Original Message-----
From:	Oyvind Teig [mailto:Oyvind.Teig@xxxxxxxxxxxx]
Sent:	donderdag 25 maart 1999 8:36
To:	Gerald Hilderink; java-threads@xxxxxxxxx; occam-com@xxxxxxxxx
Subject:	Re:It was a stressing weekend!


Good to hear about JCSP and CTJ.

Yes, we do need a real-time CSP design patterns book.

Lea's book really i a patterns book, but it's NOT CSP.

Ball, Crawford, Dr.Dobb's Sept.98. "Static fields provide one
approach" makes channels with static fields and define what they
call a Monostate Pattern. Later, in April99 they define a
"Shared Property Pattern" in the article "Are Java Applet
Independent Programs?" where they use strange or
pathological facets of the language, ask us to use it, and
call it a pattern.

"Patterns in Java," Mark Grand has a separate chapter on
real-time patterns. I think I have proved to him that the
Balking example there is not correct. The real-time patterns there
are really for beginners.

Bruce Powell Douglass, I_Logix was writing on what seemed like a
very interesting book about real-time patterns. It used to be at
www.ilogix.com/fs-papers.htm but that paper isn't there any more.
Can't find the book at amazon, either. I'm attaching a summary that
I made.


I am thinking, what if all this thoughtwork could have been
done on not reinventing the wheel, just for a cup of coffee
with the rest..

Have you, implementers, learnt anything from these JCSP and CTJ
experiences that you would want retrofit back onto occam?
Surely, occam can't be that good? Do we need a fresh brew of
an occam compiler that could generate byte-codes?