[MUD-Dev] Re: META: who are we?

James Wilson jwilson at rochester.rr.com
Tue Sep 22 19:51:20 New Zealand Standard Time 1998


On Tue, 22 Sep 1998, J C Lawrence wrote:
>On Tue, 22 Sep 1998 23:42:05 -0700 
>Richard Woolcock<KaVir at dial.pipex.com> wrote:
>> I just wish we could open a few more technical discussions.  

nobody's stopping you. ;)

>Ditto.  I would that I had the time right now.  My old question
>concerning lightweight locking supports is still out there
>(http://www.kanga.nu/~petidomo/lists/mud-dev/1998Q3/msg00467.html).
>I haven't had time to get much beyond that.

I also would be quite interested in seeing a solution to this. I'm a bit
confused about your alternative to posix semaphores (I know you said
it wasn't satisfactory anyways, but I'm still curious). Are the bits used 
like mutexes, where threads block on them until they're acquired? If
so, how did you avoid the race condition between freeing the semaphore
and blocking on the mutex?

And why do you need so many locks? I assume that  characteristic 1) is
proposed in order to allow you to construct them  on an as-needed basis. 
Why can't you allocate a fixed set of semaphores and dynamically
associate  them with resources as needed? (I believe Java does something 
of the sort with monitors.) The overall limit on the number of semaphores,
and their initialization cost, don't seem to necessarily rule this out.

James, who wants the list to be a forum for experts who are not terribly
impatient with questions asked by those who,like himself, have less
experience ;)




More information about the MUD-Dev mailing list