[Stackless] Invitation to Present at PyCon 2007

Richard Tew richard.m.tew at gmail.com
Tue Oct 31 14:35:27 CET 2006


On 10/31/06, Andrew Dalke <dalke at dalkescientific.com> wrote:
> I am planning to go to PyCon.  I think Stackless needs more exposure.
> I've given various Py* talks before.
>
> I don't know enough, yet, to be able to give a new talk.  At best it
> would be an intro talk describing what Stackless does and how to
> use it in different circumstances.  I know the tutorial talks about
> how to use Stackless with a game, but I've never done game design.
> (I have worked on numerical simulations of biomolecules.  That's
> not something where Stackless would be appropriate.)
>
> I had thoughts by now to work more on the monkeypatch idea from
> last August, but work intervened.  The proposal is due in about
> 12 hours and I don't think I can be done with a monkeypatch by then.  ;)
>
> Would something like the following be good?
>    - stackless does lightweight threading
>    - some fundamentals
>    - using stackless for boss/worker model
>        - maybe implement a turing machine?  MARS/Redcode from the 1980s?
>            something with a model and with visualization of the model
>    - doing stackless I/O
>    - .. maybe something workflow related?

What you say in the follow up mail, getting people to understand
the base reason why stackless is interesting, as I see the
above covering, seems to me to be a great topic.  I think you should
go for it.

> I'm sure I can talk about everything but the last one.  For that
> there are too many uncertainties to make something definite now.

I was thinking before you posted about doing a talk yourself,
that if a talk was to be done and aimed at getting people
aware of the benefits of Stackless, then in any ideal world
the monkey patching would be done by then removing
any concept at having to care about the problems with
synchronous IO.  But a pointer to the stackless socket
module and a clear note about the problem should be more
than enough if this is not the case.

While progress is glacial on the Windows IO completion
port based file object replacement I am writing, the hardest
part is done and reading at least can be done using channels
and asynchronous IO in the background. The hardest part
being me learning how to use IOCP with ctypes.

In an ideal world, I would push for core Python support for
cross-platform asynchronous IO, which at the least would
make it easy to use the existing objects but hook in
stackless channels, tasklets or whatever.  And I have
some thought that with generator coroutines, perhaps there
is a need for the replacement of otherwise synchronous IO
in this manner.  I know I have heard some reference to this
problem by someone using generators, in amongst the other
limitations which people using the poor mans Stackless
which is generator coroutines seem to complain about.

In any case, I say go for it, sounds great.  And let me know if
there is anything I can do, even if it is just getting my arse into
gear and getting as much of the monkey patching done.

Richard.

_______________________________________________
Stackless mailing list
Stackless at stackless.com
http://www.stackless.com/mailman/listinfo/stackless



More information about the Stackless mailing list