Hi Richard: Some comments:
>From talking to folks at EuroPython - the separate binary is a big impediment. Christian was working on integrating Stackless with Psyco and making it a module. I am increasingly convinced that this is the way to go. I was thinking about learning about stackless through looking at the stackless transform in PyPy. I am interested integrating stackless with the JIT. However it seems the knowledge to do this in PyPy is being lost. Maybe it is better than I look at Psyco instead? In order to do this, I and others would have to be so much, not spoonfeed, but taken by the hand for awhile. >At some point I decided that I wanted to have a monkey-patched socket >object, that looked and worked exactly like the standard Python socket >object, but worked with Stackless. ... >Take a look at gevent for an example of how far this can be taken: > http://www.gevent.org/gevent.monkey.html >To some degree, the code from gevent may be able to be abstracted so >that it can serve both projects. But gevent is based on libevent, and >this is a dependency that should be able to be wholly or partially >done away with. It should be possible to write this support using >standard library resources, whether asyncore or ctypes. I met Denis at EuroPython. Unfortunately I didn't get a chance to look at gEvent in detail so I couldn't ask pointed questions. As for libevent. Go uses libevent under the hood for its networking needs - which is a realistic solution. So maybe standardizing on this isn't so bad. I feel users don't want so much flexible choices but a firm networking solution. That is why stacklessocket is used - it solves problems. Myself, I use Twisted because I know how to integrate them and I really don't want to deal with low level details. Given what I have seen with generators, Stackless is far, far better.... Cheers, Andrew _______________________________________________ Stackless mailing list [email protected] http://www.stackless.com/mailman/listinfo/stackless
