A recent editorial in Nature, and a longer article by Darrell Ince, Les Hatton, and John Graham-Cumming titled “The case for open computer programs“, are just two signs of the growing pressure to raise standards around computational work in science. We can debate the exact meaning and value of reproducibility, but there’s no arguing with the fact that if scientists want to do better work, they’ll need better skills. That’s the real long-term goal of Software Carpentry.
We have finalized our first set of Software Carpentry badges—with luck, we’ll list the first set of recipients later this week.
 |
for people who complete the core curriculum |
 |
for helping to teach (either workshops or online) |
 |
for organizing workshops |
 |
for creating (or fixing) content |
It’s been a frustrating couple of days. To recap, I want to convert our material from PowerPoint to HTML5 to make it easier for people to fork and merge, to make things easier to re-style, because it’s an open format, and so on. David Seifried has welded an HTML5 audio player to Caleb Troughton’s deck.js to create a display tool, which I’m very pleased with, but the content is killing me. Seriously. Here are three slides taken from our first episode on the Unix shell:
Last Sunday, Tommy Guy and Katy Huff flew to Trieste, a small city in northeastern Italy, to assist in teaching an Advanced School on Scientific Software Development at the International Center for Theoretical Physics (ICTP).
Stefano Cozzini, Graziano Giuliani, and Antun Balaz invited us to help them teach this two week workshop which is one in a series of workshops focused on high performance computing for scientific applications. The ICTP mission is to extend access to advanced scientific tools and education to scientists from developing countries. Undergraduate students, graduate students, and post-graduate scientists gathered in Italy from nations all over South America, Asia, the Middle East, and Africa.
For the past four weeks, I’ve been meeting online with learners from the Space Telescope Science Institute to work through some Python topics and exercises. (We split the group in two for a couple of those sessions to accommodate different interests and levels.) I think it’s been going pretty well: about half the people who took part in the two-day workshop in January are still with us, and we’ve just covered basic image processing with PIL. I’m looking forward to finding out whether the model scales to more groups with more people in the coming four months.
Software Carpentry has been an open project since 2004: MIT License for the code, Creative Commons for everything else. All our stuff has lived in a public Subversion repository since then as well—it’s at http://svn.software-carpentry.org/swc if you want to check it out. Today’s question is, should we move that repo off our server and onto GitHub, BitBucket, or some other repo-hosting service? We want to make it easy for people to remix our content; as I said back in December, that means making it easy for them to fork, merge, and share. Would you be more likely to do this if our slides, learning plans, diagrams, and code samples were on some well-known hosting site rather than in Subversion on our own machine?
A dozen people have come forward since I asked last week for volunteers to make short screencasts showing how they program. I just sent them a sample problem to work on to test things out (see below the fold); the videos they create won’t be made public, but I hope it gives readers an idea of the scale of problems we’re going to be looking at. If you have suggestions for interesting problems of a similar size, please add them as comments on this post.
Cameron Neylon recently posted an article titled “Github for science? Shouldn’t we perhaps build TCP/IP first?” His argument is that the web’s a good way to move text around, because it was built by programmers, and programmers work with text. It’s not (yet) well suited to moving science around, because we don’t (yet) have something as granular and portable as text for scientific ideas. Yes, any particular piece of research can be represented as text, but so can any image, or any audio stream, or anything else—it’s the structure that adds meaning, and we haven’t (yet) agreed on structures. Circling back to today’s first post, part of what we’re trying to do is give scientists the background to understand and take part in conversations like these…
Before writing yesterday’s post about assessment, I should have explained what I mean by”fundamental concepts”. I’ll start with Lewis Epstein’s wonderful book Thinking Physics: