On 04/02/14 13:53, Kevin Conway wrote:
On 2/4/14 12:07 PM, "victor stinner"<victor.stin...@enovance.com> wrote:
>The purpose of replacing eventlet with asyncio is to get a well defined
>control flow, no more surprising task switching at random points.
I disagree with this. Eventlet and gevent yield the execution context
anytime an IO call is made or the 'sleep()' function is called explicitly.
The order in which greenthreads grain execution context is deterministic
even if not entirely obvious. There is no context switching at random.
This is technically correct of course, but in reality there's no way to
know whether a particular piece of code is safe from context switches
unless you have the entire codebase of the program and all of its
libraries in your head at the same time. So no, it's not *random*, but
it might as well be. And it's certainly not explicit in the way that
asyncio is explicit; it's very much implicit in other operations.
What's more is it shouldn't matter when the context switch happens. When
writing green threaded code you just pretend you have real threads and
understand that things happen in an order other than A => B => C.
If you like pretending you have real threads, you could just use Python
threads. Greenthreads exist because people don't want to deal with
actual pretend threads.
One of the great benefits of using a green thread abstraction, like
eventlet or gevent, is that it lets you write normal Python code and slap
your concurrency management over the top.
Right, it lets you do that and neglects to mention that it doesn't
actually work.
The whole premise of eventlet is that it allows you to write your code
without thinking about thread safety, except that you *do* still have to
think about thread safety. So the whole reason for its existence is to
write cheques that it can't cash. It's conceptually unsound at the most
fundamental level.
I'm not suggesting for a second that it would be an easy change - I'm
not even sure it would be a good change - but let's not kid ourselves
that everything is fine here in happy-land and there's nothing to discuss.
cheers,
Zane.
_______________________________________________
OpenStack-dev mailing list
OpenStack-dev@lists.openstack.org
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
