Subj : Re: pthread_create and memory
To   : comp.programming.threads
From : Bluejack
Date : Tue Mar 15 2005 10:19 am

On Mon, 14 Mar 2005 19:42:06 -0800, David Schwartz <davids@webmaster.com>  
wrote:

>
>> By "duplicate" I meant "reimplement". A language implementation may need
>> to repackage the OS APIs in a way suitable for that language, but it
>> should
>> not need to reimplement the basic functionality.
>
>     You miss my point. My point is that the functionality must be in the
> application code, therefore, it's redundant to implement it anywhere  
> else in
> addition.
>

In short, it sounds like you two agree (or nearly so) on the
philosophy of thread v. task, conceptually, but disagree
(vehemently) on where that distinction should be implemented.

Or... you haven't yet achieved clarity on what kinds of
distinctions are interesting to you.

This has been a fun thread to follow.

At this point, however, I see that an OS should provide
thread facilities such as: a parallel line of execution,
mechanism for inter-thread communication & locking,
dedicated memory spaces (minimum: stack, preferably:
stack & heap).

A language can take these basic tools and build support
for true Task Oriented Application Design (TOAD) by providing
a semantics for task encapsulation that offer developers
a clean, clear, robust, maximum performance mechanism
for utilizing the power of concurrent execution.

This all seems obvious to me: there is no overlap, language
and OS complement each other. Undoubtedly, I am missing
something. What?

-bluejack

PS. I am working on a paper laying out the correct principles
of TOAD, and proposing the language features required to
support it. You can understand why this has been a fascinating
thread. Carry on!

.