--- "Miller, Raul D" <[EMAIL PROTECTED]> wrote:

> Oleg Kobchenko wrote:
> > Because J instance is not re-enterable from
> > different threads, as shown, we need different J
> > instances, one per thread.
> 
> Agreed.
> 
> > To start a thread, we need a callback: different
> > callbacks (call addresses) for different thread-instances.
> 
> I disagree.
> 
> To start a new thread in an existing process, something
> like this might be necessary.  However, I believe J
> currently is designed around 1 J thread per process.
> I believe that many aspects of the existing J implementations
> depend on this design.

I believe this not true, or it will be a big disappointment.
I always assumed that you can run several J instances
on different threads simultaneously. One big example is
ASP and ASP.NET: I tested it and it worked. Another example
is my jasync (async.ijs), which with a help of 
external callback, encapslating J instance on its stack, 
is also able to do it.

(All with reasonable concurrency guards like mutexes
for file, shared mapped vars etc. operations.)

> Thus, to start a new J thread (control stack), you need
> to start a J instance, which in turn means starting a
> new process (independent writable virtual memory representation
> of the machine).

Different processes is outside of this scope. With the
assumtion above, the problem is to find a solution for
the cdcb problem.


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to