Thread Synchronization (Was Re: Sockets vs. net connection)

From: William Kempf (sirwillard_at_[hidden])
Date: 2001-01-04 10:17:18

--- In boost_at_[hidden], "Milutin Jovanovic" <miki_at_o...> wrote:
> I am suffering extreme time shortage, so I have kept my mouth 
closed for a
> very long time. However, at this stage I have to ask the following. 
What is
> the aim of the threading support we are planning? Is it:
> 
>  a) rudimentary support for the multi-threading promitives
>  b) provinding a complete set of sophisticated objects to support
> multi-threading
> 
> A) means providing very basic mutex, conditional variables, thread, 
all
> independent of each other. B) would provide these and more, and 
allow more
> sophisticated uses like waiting for multiple conditions, prioritised
> mutexes/CV's, even allowing to wait on socket (and any other 
convinient
> condition) together with threading primitives.

It's a multi-staged plan.  B) is going to require A) to be 
implemented, and frankly, no one really knows what B) should be 
(researchers have been pondering this for decades, and though we've 
got some better constructs, even these are full of problematic 
cases).  So the goal is to initially provide A), meaning only the 
basic primitives implemented portably, correctly, safely and simply.  
From there programmers will have a common base to work out from in 
trying to define the next generation of threading concepts (including 
those already available in other languages, such as CSP).
 
> I for one have no preference on the choice. But every time people 
say "mutex
> should be independent of thread" or similar, my question pops up in 
my mind
> again.

A Mutex concept, though related, should always be independent of the 
Thread concept.

Bill Kempf 
> Cheers,
> 
> Miki Jovanovic.