On Wednesday, June 27, 2012, Christian Tismer <[email protected]> wrote: > Hi Richard, > > (why do you reply in private?)
Accident. > > On 26.06.12 21:59, Richard Tew wrote: >> >> On Wed, Jun 27, 2012 at 1:16 AM, Christian Tismer <[email protected]> wrote: >>> >>> Stackless changes the python executable and is therefore not >>> absolutely easy to install. >>> This is the reason why the greenlet is preferred so often over stackless. >> >> I would say it was part of the reason. The other part is that it >> tends to come as part of a library that enables easy use of >> asynchronous IO, like gevent or eventlet. > > Well, a recurrence. eventlet was probably developed on greenlet because of > this difficulties. >> >> We should also concentrate on doing whatever needs doing on >> stacklesslib and consider adding that as part of the Stackless >> distribution. > > If you can tell me more on this, that would be great. > I need to prepare my EuroPython talk where such things > should be mentioned. Kristján and I have discussed this in the past. Like eventlet and gevent, it offers microthread friendly abstractions for IO. And monkeypatching them in place so than existing threading-based code can transparently just work with the scheduler. But there's the question, is it better to keep this a pypi hosted and installable package, or to provide it out of the box as part of the SLP distribution. Maybe it is not important. >> >>> It suddenly struck me that the installation problem might simply >>> vanish or become invisible? >>> With virtualenv or maybe some modification, it should be not so hard >>> to make a stackless installation that creates a virtualenv configuration >>> that does not impose headaches for newcomers. >>> >>> My dream-solution is >>> >>> pip install stackless >> >> I think it is definitely a good idea, but it seems complicated. >> >> We would be replacing the Python runtime, and injecting some standard >> library module changes. If someone is using Python 3.2.1, I don't >> think it is acceptable to silently give them the Stackless Python >> 3.2.2 runtime. >> > Yes, this needs a bit of thought. > Actually, we could require a compatible version and otherwise reject > installation. To ask for the latest update of a python 2.x is probably > ok. Alternatively, to support 2.x.y where y <> latest is possible, but > probably no good idea. > > In any case, if people don't like what they get, they can pip uninstall, > and be back where they were before. I think this makes a huge difference. > > I think it makes sense to play even a bit more with the idea of > "treating" stackless like an extension without actually being one. > It might even be possible without virtualenv if we can manage to find > a good bootstrap trick. virtualenv was the vehicle to think about it. > Can be, can be not the best solution. Anyway better than all the humungous > hackery that I thought or two. > > At least for python 3.3, I think something is possible using the builtin > venv features, installing stackless bootstrapping in a user environment. > > If you get a concrete idea, please tell me ASAP, because I want to draw > a road-map where stackless will probably go in the near future. > My EuroPython talk is early next week, I'm preparing till Saturday. Will do. Cheers, Richard. >
_______________________________________________ Stackless mailing list [email protected] http://www.stackless.com/mailman/listinfo/stackless
