Addition of multiprocessing ill-advised?
jnoller at gmail.com
Wed Jan 28 14:49:28 CET 2009
On Wed, Jan 28, 2009 at 8:32 AM, Steve Holden <steve at holdenweb.com> wrote:
>> I have found no problems with it - I've recently integrated it with my
>> event/component framework (1). In my library I use Process, Pipe
>> and Value.
> It will be a great library in time, but the code was immature and
> insufficiently tested before the 2.6 release. The decision to include it
> late in the release cycle
> There are 32 outstanding issues on multiprocessing, two of them critical
> and four high. Many of them are platform-specific, so if they don't hit
> your platform you won't mind.
See my reply to the thread I just sent out; I don't disagree with you.
However, there are not 32 open bugs:
Man, I hope that url comes through. As of this writing, there are 18
open bugs assigned to me for resolution. Of those, 2 are critical, but
should not be - one is an enhancement I am on the fence about, and one
is for platforms which have issues with semaphore support.
However, I agree that there are bugs, and there will continue to be
bugs. I think the quality has greatly increased since the port to core
started, and we did find bugs in core as well. I also think it is more
than ready for use now.
> Jesse did a great job in the time available. It would have been more
> sensible to wait until 2.7 to include it in the library, IMHO, or make
> the decision to include it in 2.6 in a more timely fashion. The one
> advantage of the inclusion is that the issues have been raised now, so
> as long as maintenance continues the next round will be better.
Again, I don't disagree. Alas, the PEP resolution and proposal was
greatly delayed due to, well, someone paying me money to do something
else ;) - that being said, yes, the decision was made late in the
game, and was disruptive.
Maintenance is going to continue as long as I continue to have an
internet connection. Heck, there are enhancements to it I really want
to add, but I swore off those until the bugs are closed/resolved and
my pycon talks are in the can.
More information about the Python-list