[IPython-dev] Starting to plan for 0.11 (this time for real)
fperez.net at gmail.com
Thu Oct 28 20:01:11 EDT 2010
I know we've said several times that we should release 0.11 'soon', so
I forgive anyone for laughing at this email. But don't ignore it,
this time we mean it :)
We now have a massive amount of new code in the pipeline, and it's
really high time we start getting this out in the hands of users
beyond those willing to run from a git HEAD. 0.11 will be a 'tech
preview' release, especially because the situation with regards to the
parallel code is a bit messy right now. But we shouldn't wait for too
Brian and I tried to compile a list of the main things that need work
before we can make a release, and this is our best estimate right
- Unicode: this is totally broken and totally unacceptable. But I'm
pretty sure with a few clean hours I can get it done. It's not super
hard, just detail-oriented work that I need a quiet block of time to
- Updating top-level entry points to use the new config system,
especially the Qt console. Brian said he could tackle this one.
- Final checks on the state of the GUI/event loop support. Things are
looking fairly good from usage, but we have concerns that there may
still be problems lurking just beneath the surface.
- Continue/finish the displayhook discussion: we're well on our way on
this, we just need to finish it up. We mark it here because it's an
important part of the api and a good test case for how we want to
expose this kind of functionality.
- Move all payloads to pub channel. This is also a big api item that
affects all clients, so we might as well get it right from the start.
I can try to work on this.
- James' web frontend: I'd really like to get that code in for early
battle-testing, even though it's clear it's early functionality
subject still to much evolution.
That's all I have in my list. Anything else you can all think of?
As for non-blockers, we have:
- the parallel code is not in a good situation right now: we have a
few regressions re. the Twisted 0.10.1 code (e.g. the SGE code isn't
ported yet), the Twisted winhpc scheduler is only in 0.11, and while
the new zmq tools are looking great, they are NOT production-ready
quite yet. In summary, we'll have to warn in bright, blinking pink
letters 1995-style, everyone who uses the parallel code in production
systems to stick with the 0.10 series for a little longer. Annoying,
yes, but unfortunately such is life.
- our docs have unfortunately gone fairly stale in a few places. We
have no docs for the new Qt console and a lot of information is partly
or completely stale. This is an area where volunteers could make a
huge difference: any help here has a big impact in letting the project
better serve users, and doc pull requests are likely to be reviewed
very quickly. Additionally, you don't need to know too much about the
code's intimate details to help with documenting the user-facing
Plan: I'd love to get 0.11 out in the first week of December. John
Hunter, Stefan van der Walt and I (all three contributors) will be at
Scipy India in Hyderabad Dec 11-18, and there will be sprint time
there. Ideally, we'd have a stable release out for potential sprint
participants who want to hack on IPython to work from. It would also
be a good way to wrap up a great year of development and de-stagnation
of IPython, leaving us with a nice fresh ball of warm code to play
with over the winter holidays.
More information about the IPython-dev