← Back to team overview

nova team mailing list archive

Re: twisted v. eventlet


I ended up giving an impromptu lightning talk on async network programming
to some folks at local python user group. I introduced twisted via some
examples and then eventlet via the same example. It was clear from the
audience that greenlet style async programming seemed significantly more

I realized another significant point in favor of eventlet, is the ability to
utilize pdb in application code. In twisted pdb is fairly useless, in that
any stack trace is shallow to the reactor across a deferred. With eventlet
the entire greenlet callstack is still maintained across switches and
introspectable from pdb.

So fwiw, i've come around from leaning towards twisted, to thinking that a
greenlet approach is probably a good way to go.  What's a little more
unclear is what advantage eventlet has over gevent?



On Tue, Aug 3, 2010 at 10:26 AM, Kapil Thangavelu <kapil.foss@xxxxxxxxx>wrote:

> On Tue, Aug 3, 2010 at 9:13 AM, Ed Leafe <ed@xxxxxxxxx> wrote:
>>> On Aug 3, 2010, at 7:01 AM, Kapil Thangavelu wrote:
>>> > Looking at mailing list traffic alone, the twisted community still
>>> dwarfs significantly the eventlet community.
>>>         Probably not a very good comparison. Eventlet is pretty much
>>> based on Stackless, and implements the Stackless changes into a CPython
>>> module. So a better comparison might be the Stackless + eventlet
>>> communities.
>>> -- Ed Leafe
> I'm not sure that's valid given the specific references to the stackless
> implementation on a large portion of the stacklist mailing list subject
> lines, but even stackless+eventless combined is only 85% of the twisted list
> alone (ignoring the twisted-web list) for mailing list traffic for 2010.
> However these measures are rather crude (based on mailman mbox size), but I
> put together a quick spreadsheet for anyone who wants to check my math :-)
> http://spreadsheets.google.com/ccc?key=0Ar5R--MATn3QdGdsN1kyd3NPVUxKY19BX0M1ODNsakE&hl=en
> i find these measures to be irrelevant compared to feature (or misfeature
> comparisons), but I at least wanted to point out that the twisted community
> is still quite vibrant 10 years running.
> cheers,
> Kapil