RFC: Multithreading design constraints

From: scleary_at_[hidden]
Date: 2000-06-07 09:06:30

> One thing that noone has mentioned is process/task related issues.  You 
> can, on Win32 for example, implement synchronization primitives that are 
> fine within process, but will break out-of-process.  I am assuming that 
> all of the semantic guarantees made so far relate to in-process 
> concurrency (and may, but not necessarily, extend out-of-process).

The semantic guarantees scale seamlessly to process synchronization.  Just
allow each object to have a name.

About 1 1/2 years ago, I began playing with this dimension of the
synchronization game.  There are thread-level sync. primitives and
process-level sync. primitives.  I actually wanted to go further -
synchronize processes running on different computers on a network --
essentially, computer-level sync. primitives.

The semantic guarantees will also extend to computer synchronization
primitives -- if we add another name, the computer name.  On that computer a
"synchronization server" daemon/service will be running that provides the
computer-level synchronization primitives.  I actually had a simple form of
this running at one time.

Let's not look at this for now -- we're going to have a hard enough time
with thread/process level!  But keep this avenue open for future expansion.
As distributed systems become more common, something like this may need to
develop.

        -Steve