[No monthly wrap-up for now; other things to talk about..]

At the Python in Astronomy workshop, we had a discussion about career credit for software development, the notes of which are here. During the session I fired off a Tweet to C. Titus Brown to ask for ideas. This spawned a Twitter discussion (which ensued during the face-to-face discussion!). Titus wrote a blog post which spawned more interesting discussion, both in the comments and in follow-up posts:

A lot of the commentary so far has been from the point of view of biology or particle physics, and I would argue that observational astrophysics is in a middle ground between these two. We don’t have the everyone-has-their-own equipment like biologists now do we have the everyone-uses-one-big-facility like particle physics; instead we get our data from a mix of observatories which don’t belong exclusively to anyone. Our data is all digital; we need software to analyze it, much of which is not specific to any one facility. To avoid duplicating effort, it makes sense to develop a community software library that everyone can use and contribute to. Of course, one such library is astropy, but it’s not the only one.

Should the work that goes into developing software-for-science be called science itself? I think that it’s more like engineering, but with the critical difference that it has to be done by scientists or at least people with a deep connection to, or understanding of, the science involved. In astronomy, with its lack of commercial applications, there’s unlikely to ever be an off-the-shelf product that will serve our needs; if we don’t write our own software, no one will do it for us. Writing the code isn’t doing science per se, but it’s enabling science, in much the same way as building instruments for telescopes is. [I think a different argument might possibly be made for computational astrophysics, but that’s someone else’s to make.] Both are an important part of the research ecosystem. In some ways they function like the “service” part of academia: necessary for everyone but performed only by some.

So back to what we were discussing at Python in Astronomy: how to credit people for their contributions to software? Right before our workshop, a report on tthe topic came out; it makes some interesting suggestions but a lot of them focus on citations. For software, the citation system seems clunky: authors of a paper are fixed, but the contributors and maintainers of software can change. Conference presentations don’t seem the right avenue either. One of the things we talked about during our discussion was how to create metrics for software/contributions; openhub.net seems to have made a start in that direction.

This is an important discussion that is just getting started. I don’t see a lot of it happening in Canada just yet (maybe I’m not looking in the right places?) I suspect we will probably end up doing what we usually do and following the bigger players, but I think we better pay attention. We can do better research and learn more about the world if we’re not all reinventing the wheel, and we need to work out how best to reward people who make the wheels better.