[AstroPy] Co-ordinating Python astronomy libraries?
Perry Greenfield
perry at stsci.edu
Tue Jul 6 17:03:47 EDT 2010
I'm heartened by all the responses so far in seeingg that there is
interest at some level in the topic.
We don't expect magic to happen and suddenly see coordination happen
painlessly, nor simultaneously on all fronts. I (I'm speaking for
myself on this) don't think that it is bad that there are multiple
approaches tried for various things. Multiple approaches often show
good ideas in each approach that others can learn from.
Nevertheless, I can't but feel that in some areas the duplication of
effort has gone too far. One example would be for FITS WCS modules. It
would be good to understand if we could merge some of these efforts
into a smaller number, or at least attempt to. There seem to be
multiple parallel efforts to replicate the IDL astron library that may
benefit from sharing of work and coordination.
At this point it seems clear sphinx is going to be the documentation
standard, and we might as all well begin using it or converting to it
(we at STScI are in the process of doing that).
We will be setting up the public astrolib repository this week (and
I'll send an email when it is available). Are there any volunteers
interested in migrating to it? It probably would be good only to have
1 or 2 to start with to see what issues arise.
For those reluctant, we are curious about what the main obstacles are.
E.g.,
1) SVN? (what do you use?)
2) trac? (what do you use?)
3) security? (e.g., having other developers having the ability to
commit changes to your code)
4) government regulations or laws (e.g., I have heard some are
prevented from using some US hosts because of US laws)
5) worries about the safety of the code (we back up the repository
nightly to disks at STScI)
As to 3) I think it is reasonable to list those who are responsible
for making decisions about what changes are accepted for any
particular module (i.e., local control), and should anyone abuse that
by repeatedly committing changes that aren't approved by the lead or
leads on that module, they would lose their commit privileges. In
other words, by moving your code to the astrolib repository, you
aren't ceding control over changes to that code.
Any other thoughts?
Perry
More information about the AstroPy
mailing list