[Numpy-discussion] What is consensus anyway

Chris Barker chris.barker@noaa....
Mon Apr 23 17:46:18 CDT 2012

On Mon, Apr 23, 2012 at 3:08 PM, Travis Oliphant <travis@continuum.io> wrote:
> Right now we are trying to balance difficult things:  stable releases with experimental development.

Perhaps a more formal "development release" system could help here.
IIUC, numpy pretty much has two things: the latest release (and past
ones) and master (and assorted experimentla branches). If someone
develops a new feature, we can either:

have them submit a pull request, and people with the where-with-all
can pull it, compile, it, and start tesing it on their own -- hsitory
shows that this is a small group.

merge it with master -- and hope it gets the testing is should before
it becomes part of a release, but: we are rightly heistant to put
experimental stuff in master, and it really dont' get that much
testing -- again only folks that are building master will even see it.

Some projects have a more format "development release" system.
wxPython, for instance has had for years development releases with odd
numbers -- right now, the official release is 2.8.*, but there is a
2.9.* out there that is getting some use and testing. A couple of
things help make this work:

1) Robin makes the effort to put out binaries for development releases
-- it's easy to go get and give it a try.

2) there is the wxversion system that makes it easy to install a new
versin of wx, and easily switch between them (it's actually broken on
OS-X right now --- :-) ) -- this pre-dated virtualenv and friends,
maybe virtualenv is enough for this now.

Anyway, it's a thought -- I think some more rea-world use of new
features before a real commitment to adopting them would be great.



Christopher Barker, Ph.D.

Emergency Response Division
NOAA/NOS/OR&R            (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception


More information about the NumPy-Discussion mailing list