On Wed, Sep 19, 2001 at 08:35:40PM -0700, Ryan Bloom wrote:
> I agree that having them inconsistent is incredibly silly. I am wondering
> why you think the 1.3 one makes more sense. I have told you why I think
> calling out that it is a process makes more sense to me. I am trying to
> understand your point of view.
Since all of the other alternatives (fcntl, flock, etc) are also
process-level, it follows that any pthread accept mutex should be
in a similar vein. It wouldn't make sense to have an intraprocess
lock option when all of the others are inherently interprocess
locks. (Remember that fcntl can't be used to lock within a single
process.) Not to mention that we, as a group, don't like single
process MPMs - so having the AcceptMutex be an intraprocess lock
isn't really feasible as a *common* MPM option. Therefore,
I think there is no need to make the unnecessary distinction - the
context in which this mutex is used indicates that it should be an
interprocess lock.
It is something that I can go either way on - I like pthread
better. But, we should definitely be consistent. -- justin
