[IPython-dev] Starting to plan for 0.11 (this time for real)

[Fernando Perez]

On Sat, Oct 30, 2010 at 5:55 PM, Erik Tollerud <erik.tollerud at gmail.com> wrote:
> +1 on removing Twisted (mostly from a user perspective) - I've had a
> number of different strange dependency problems with Twisted while
> trying to get IPython working on a cluster, so ditching it entirely
> would make such things easier for experimenting with .11

Thanks for the feedback.

> With that in mind, I'm in the early stages of developing a small
> application intended for use on the aforementioned cluster.  I'd like
> to write it for use with .11 instead of .10 if possible (thinking of
> it partly as a test case to help with debugging .11).  Because of the
> network setup and the possible need for remote collaborator input, the
> ideal situation would be to use the html frontend to get at the
> cluster.  But that's a non-starter if there's no authentication - is
> it at all realistic to try for the html frontend with authentication
> in .11 (presumably w/ Tornado, based on the above discussion)?

That's our hope, but obviously I can't promise anything :)  But yes,
that's one big reason to think of Tornado, so that we can at least
have basic auth/ssl support for the web entry point.  The kernels by
default only listen on localhost, so it's not that bad: you have a
problem with being open to other local users, but not to the whole
internet.  But since the webnb effectively forwards the
obscure/multiport/local zmq connections over nice, clean,
everybody-understands-it http, it all of a sudden turns an improbable
problem into a likely catastrophe.

Let's see how the next few weeks go, and all help on this front will
be very welcome.

Cheers,

f