Daily Life in an Ivory Basement

After my recent next-gen sequencing course, which was supposed to tie into the whole software carpentry (SWC) effort but didn't really succeed in doing so the first time through, I started thinking about the Right Way to tie in the SWC material. In particular, how do you both motivate scientists to look at the SWC material, and (re)direct people to the appropriate places?

It's not clear that a Plan is in place. Greg Wilson seems to assume that scientists will find at least some of the material immediately obviously usable, but I think he's targetted at a more sophisticated population of users -- physicists and the like. My experience with bioinformaticians, however, is that they either come from straight biology backgrounds (with little or no computational background and rather limited on-the-job training), straight computation backgrounds (with very little biology), or physics (gonzo programming skills, but no biology). The latter fit neatly into the SWC fold, but they (we ;) are rare in biology. I think computer scientists and biologists are going to need guidance to dive into SWC at an early enough time for it to be the most rewarding.

So, what's a good model for SWC to guide scientists from multiple disciplines into the appropriate material? It's obviously not going to be possible to have Greg et al. tailor the SWC material to individual subgroups -- he doesn't know much (any ;) biology, for example. I don't have the time, patience, or skillset to integrate my next-gen notes into his SWC material, either. So, instead, I propose the hub & spokes model!

Here, the "hub" is the SWC material, and the spokes are all of the individual disciplines.

Basically, the idea is that individual sites (like my own ANGUS site on next-gen sequencing, http://ged.msu.edu/angus/) will develop their own field-specific content, and then link from that content into the SWC notes. This way the experts with feet in both fields can link appropriately, and Greg only has to worry about making the central content general -- which he's already doing quite well, I think. Yes, It's more work than asking Greg to do it, but frankly I'm going to be happy with a kick-ass central SWC site to which I can link -- right now it's dismayingly challenging to teach students why this stuff matters and how to learn it.

From the psychosocial perspective, it's a great fit. Students can get hands on tutorials on how to do X, Y, and Z in their own field -- and then connect into the SWC material to learn the background, or additional computational techniques in support of it. Motivation first!

What do we need SWC to do to support this? Not much -- basically, the central SWC notes need to be stable enough (with permalinks) that I can link into them from my own site(s) and not have to worry about the links becoming broken or (worse) silently migrating in topic. There are other solutions (wholesale incorporation of SWC into my own notes, for example) but I think the permalink idea is the most straightforward. Oh, and we should have a Greg-gets-hit-by-a-bus plan, too; at some point he's going to move on from SWC (perhaps when his lovely wife decide she's had enough and he needs to stop obsessing over it, or perhaps under more dire circumstances ;( and it would be good to know who holds the domain and site keys.

Thoughts? Comments?

--titus

The last two weeks were pretty miserable, for some scientific/collaboration reasons as well as some personal reasons (visiting sick parents != fun). Two things that weren't miserable -- that were in fact quite fun -- were PyOhio and the Science 2.0 talks in Toronto.

PyOhio was a nice little community-based conference, and I got a chance to give my perspective on a alternative to buildbot, as well as Snakebite and the Python Buildhaus; you can watch the recording if you want details. I'll be posting more about pony-build (the aforementioned buildbot alternative) as things progress; I'll just say that I have now fired it "in anger" and I'm working on a release.

PyOhio was well-run and had a wide attendance, although attendance waned by the end of the second day. I'd suggest that they do a single track with 30 minute talks next year, as the 1 hr talk format left some people flailing for material, and others rambled. We stayed at a nearby Red Roof Inn, which wasn't unpleasant but was not exactly wonderful, either.

One of the highlights of PyOhio for me was the opportunity to bring two students along from MSU, Owen and Eric. I think they would have been a bit more wary of the crowds at PyCon; PyOhio was a gentle introduction to the very nice Python community.

The following week, I navigated the passport renewal process in Detroit and took VIA Rail (the Canadian train company) from Windsor to Toronto. The trip itself was awesome -- business class: meals + coffee served, wifi, and power -- and gave me a chance to rewrite my talk in a better direction. The following morning I attended the UToronto CS dept summer projects demo session which was stunning, and then gave my talk at the Science 2.0 meeting. Others have covered that meeting in detail -- in particular, see Steve Easterbrook's liveblogging, which makes me seem more coherent than I was, and of course the Twitter feed -- so I'll just say that I felt quite at home with the other speakers. It was great to meet Jon Udell, Cameron Neylon, and Michael Nielson; Victoria Stodden blew me away with some actual research! numbers! on the problem of openness and reproducibility. I will be paying attention to her work in the future and making use of her publications in my classes.

My talk was OK; I felt like it was a rather lightweight collection of opinions, however justifiable. You can judge for yourself: Jon Udell blogged the whole thing in great detail.

Probably the best one-sentence summary of the event was given by Milan Davidovic:

The main message today seemed to me to be: get your work out there (i.e. on the Web), open it up to collaboration, and stop worrying so much about your precious IP; those who play their cards too close to the chest will be left behind.

Yep. Nail on the head.

I gather that videos and slides will be posted soon, if they're not already up somewhere; Google is your friend.

The whole event made me realize that while I want to be a leader in the Science 2.0 arena, I want to lead by example and not by giving lots of talks. I'm much more comfortable with data than opinions, despite my tendency to be heavier on opinions these days. It also renewed my determination to blog: hence this post ;).

After making it back down from Toronto, I then collapsed into a small ball and slept off the two weeks. (j/k: I came into work and had to deal with Issues.)

--titus

I recently gave an informal talk on Software Carpentry for the Caltech e-Science 101 course. Since even "Intro Software Carpentry" is a whole course of study, I obviously couldn't cover much, but I tried to motivate people to get interested. And, of course, I pushed testing. TESTING, DAMNIT!

Anyway, here are the talk slides (in PDF format).

--titus