From: Alec Cawley <alec@cawley.demon.co.uk>
Newsgroups: comp.sys.transputer
Subject: Transputer withdrawal symptoms
Date: Tue, 29 Sep 1998 19:53:58 +0100
Organization: Lost in the Wilderness
Message-Id: <frX5fBAGzSE2EwBU@cawley.demon.co.uk>
Mime-Version: 1.0


After five years programming transputers, I am back on conventional
processors, with a conventional RTOS - and it is coming as quite a
culture shock.

I have been using C++ not occam, but with an occam-like structure - no
shared variables, communication by channels, and communication is
synchronisation. 

Now I return to the world of semaphores, mutexes, message buffers with
no ALT equivalent, and it seems so incredibly messy. I can't decide
between a task oriented design and a data oriented design - I need to
describe the data before I can describe the tasks which may access it,
but I need to describe the tasks before I can justify the data that uses
it. And how do you have a task that can respond to many stimuli without
an ALT? Aaaargh!

Of course, these are all soluble problems. But the solutions involve a
mishmash of different constructs, whereas the transputer model solved
all the variations with a single construct.

Of course, with my RTOS, I am supposed to get a nice interactive
debugger and a smart development environment of the sort I never had on
the transputer, so it may all turn our happily in the end. I hope.

Why, oh why, did inmos screw up the commercial side so badly? Bits of
the technology are *still* ahead of the rest of the world.

-- 
Alec Cawley