[AstroPy] AstroPy Digest, Vol 58, Issue 16

Erik Tollerud erik.tollerud at gmail.com
Wed Jun 15 01:06:59 EDT 2011


I mostly agree with Thomas' sentiment that it's important that the
code base itself be on neutral ground.  If nothing else, that makes it
easier to work with and manage by community members not at that
institution, and makes it more firmly community-driven. (Oh, and to be
clear, I don't mean to imply any suspicion or criticism of STScI -
this is just a general view that happens to apply in this particular
case.)

> I don't think this is quite true. From what I understand, Enthought
> does host the source code repositories for both numpy and scipy
> (scipy.org servers belong to Enthought). Why, they host this mailing
> list ;-).

Actually, all of numpy and scipy (except maybe some of the scikits?)
development is on github now.  The same is true of ipython... that's
part of the reason why my personal slant is that github is the way to
go (see below).  The web servers are hosted by Enthought though, and I
think no one disagrees that it would be great for institutions like
STScI (or individuals at those institutions) to take leading roles...
But scipy, numpy, and ipython moved to github because it's far easier
for the community to get involved using tools like those available on
that site.  Even moreso than for those packages, it's crucial that
there be a lot of community buy-in here, because of the
"roll-your-own" culture that seems to pervade much of astronomy.


> It is true that we fund the astrolib repository, and that it currently
> isn't a free hosting, it is hosted off of STScI machines. And it is a
> generic SVN repository which could easily be relocated should that
> commercial hosting service disappear. (We regularly backup the
> repository locally). And astrolib.org is a generic url with no
> intrinsic STScI link. Doing both of these was to try to avoid the
> institutional ownership issue that you raise.
>
> But we will use whatever repository system people will use the most.
> If that would be git, that would be fine with us (or me anyway :-)
> It's not clear yet to me if that is git (votes?).

And it's certainly great for the community that you're willing to do
things like this, don't get me wrong!  But the people who made github
and bitbucket have spent a long time (and a *lot* more money)
developing web sites to solve exactly the problem we're facing here,
so I think it would be a bit strange not to take advantage of it.  I
promise once you get used to git or mercurial, you won't understand
how you could ever have used SVN/CVS!

On the other hand, as Thomas pointed out in another post, I think now
may not be the time to decide this quite yet - we need a clearer
vision of the overall structure before we try to implement anything.
But certainly in the near future, a poll is in order.


-- 
Erik Tollerud



More information about the AstroPy mailing list