The Canonical model for scientific software

This is one of those blog posts that gets inspired by something I hear in a podcast. Also one of those posts that I wish there were easy ways to link to specific pieces of audio and just those pieces of audio.

Earlier this morning I was listening to episode 77 of Floss Weekly, an interview with Jono Bacon about Ubuntu. In a great podcast, one number that Jono, who is the community manager, mentioned was the number of active contributors to Ubuntu, somewhere between 50000 – 100000. Now that’s a large number, more than I had expected, and it got me thinking about the role Canonical plays in the Ubuntu world

If there is one area where I can realistically stand up with a straight face and say I have some authority it’s in scientific software. It’s what I have spent almost all my career (except for the past year) doing. One of the questions I have always asked myself is about models that result in sustainable, high quality software that can be accessible to people to do good science, and it’s been a tough question to answer as there are so many variables, so many user types and so many needs, coupled with tremendous pricing pressures and, in most cases, a small addressable market (when it comes to pharma usually a shrinking market). But the podcast really got me thinking about some possible models. Now to put some biases in play here. From my years as a product manager and subsequently observing different development models, I believe that you get most success either from small, very focussed, teams or from distributed teams (the latter would be the Ubuntu approach). For some scientific projects, I believe the latter has some interesting twists based on the Canonical model.

Now to be clear, the Linux operating system and scientific software are not the same thing, not even close. But what if there were pieces of software (R comes to mind and actually seems to use many of those principles) for which one could get a few 1000 active contributors, or a type of software for which there was a decently large user bases (many 100′s or 1000′s). What if there was an entity or entities whose role was not too dissimilar to the role Canonical has taken for Ubuntu and the goal of developing a best of breed Linux distro. The reason I say this is that a lot of good scientific software is open source, or should be, with a core group of developers and access to the wider body of scientific programmers with interest in the field. The entity can fund and focus on the key aspects, e.g. scalability, and provide value adds like customization and turnkey solutions. It could even be a non-profit organization (an idea that a former colleague was really intrigued by) focussed on solving some of the core challenges. Regardless of the business model here, the concept I am trying to focus on is the ability of an organization to facilitate and foster open source development, and build quality, sustainable software. In many science projects, there tends to be a lack of coordination, since it is managed by a single lab and a postdoc or two (this seems to be changing a little). Or you have software where for a fee you can get the source, but there is a lot of fragmentation of the code base and no sustained effort to solve problems which impact a community, since most scientists write code for parts that solve a problem particular to them. What makes this especially interesting is the rise of distributed version control systems, which make this whole process of managing source and projects, forking code, etc that much easier to handle.

I don’t think this model works for all scientific projects. In fact, just for a few. But there are enough classes of software where if you had an organization or company fostering open source software, managing and sustaining it, IMO we’d get some good quality software out there, and less fragmentation (we don’t need 100 alignment algo’s or MM apps that do exactly the same thing). How the business aspects would work is something that needs to be analyzed, but it’s not like scientific software companies are laughing their way to the bank. Already you are seeing examples of companies pushing and supporting the development of open source software systems like Agilent just did with Cytoscape (subscription required) through their philanthropic arm. If you are smart, you know that in the end that is likely to help your own cause.

Essentially what I seem to be gravitating towards is a departure from the traditional in-licensing model of scientific software companies (if you are doing your own development that falls into the small focussed effort category) and a model where the company essentially fills the role Canonical does for Ubuntu. You could have a couple of companies out there in that role, and find the appropriate balance between supporting open source and building a business, or you could go the foundation route. A topic to be re-visited.

Related articles by Zemanta

• Beyond Ubuntu: Canonical Diversifies Its Revenue Streams (slumpedoverkeyboarddead.com)