Re: JavaMemoryModel: PC language alert: more suggestions

From: Victor Luchangco (Victor.Luchangco@Sun.COM)
Date: Tue May 18 2004 - 14:32:46 EDT

Jeremy Manson wrote:
> Eliot Moss wrote:
>> How about:
>> internal action
>> (thread-) local action
>> private action
>> quiet action
>> as possibilities?
> The names you suggest don't connote the fact that the thread in question
> has to repeat infinitely.
> We've basically decided to go with "thread divergence action" (to
> distinguish it from general divergence).
> Jeremy


At this point, I'm more concerned about the concept than the name.
I still don't understand why we need to distinguish "catatonia"
actions from any other kind of internal action. Bill's last
email on this subject suggested that just modeling all local
actions could work, and that the disagreement was about what
was more natural. I think the path you are taking is a dangerous
one: there have been many years of experience with explicit
modeling of internal actions in a variety of formal models
for concurrent systems--indeed, all the formal concurrency
models I can think of offhand. The introduction of a special
kind of action seems needless to me. There may be some property
you want to capture that is not captured in a straightforward
way by modeling local actions explicitly, but I'd like to see a
concrete example. I haven't even heard a clear statement of
what that property might be--just some claim that using local
actions may be less natural. But from the discussion on the
list thus far, it sounds like Jerry, Eliot and I, at least,
prefer internal actions. (Jerry and Eliot, please let me know
if I am wrong about this.)

It would help me a lot if you gave a program that would not
have the right semantics (presumably a progress problem) if
we modeled every internal action explicitly (as a no-op for
the rest of the JMM), and required that in any infinite
execution in which no thread takes an infinite number of
external actions, every thread either terminates or takes
an infinite number of internal actions. (The latter case
corresponds to an infinite loop.)


JavaMemoryModel mailing list -

This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:01:07 EDT