Friday, February 2, 2007

The GPL Won’t Solve This NIH Problem

I almost called this “NIH — Not Just for Programmers Anymore,” “Physiologists Who Need a GPL” or perhaps “Forking the Human Body.” The 1st choice is particularly fun because NIH has a double meaning, one for the IT sector and another for health care.

There was an interesting tale of forking at this week’s DCCI conference at the National Academies of Science and Engineering. The story was told by Brian Athey, a bioinformatics professor at the University of Michigan. Because his slides are not yet posted, I need to reconstruct the story from my sketchy notes.

[Head cross-section]Back in the early 1990s, the federal National Library of Medicine began its Visible Human Project. There’s a nice article at GE explaining how the various 3D imaging technologies (plus cadaver slices) were used to generate the dataset. (For the IT types, the GE study talks about using the big iron of the day — SGI dual 150 MHz screamers). As Athey told it, after the NLM had a virtual man (and then woman), everyone else wanted one, including DARPA with its Virtual Soldier, NASA with its Virtual Astronaut, and others I couldn’t type quickly enough. Even Ray Kurzweil wants to get involved.

If I understood his point, there’s a lot of re-inventing the wheel, everyone creating their own human by starting over from scratch, rather than doing cumulative innovation where one researcher builds upon (and contributes back to) the infrastructure created by others. BTW, if you go to NLM website, their license even has a form of reciprocity.

It’s yet another reminder to the GPLniks and others who believe that compulsory sharing is the answer to all the world’s problems. Take embedded Linux. Even if you can force people to share changes, that doesn’t mean that those changes will be used: it took several years to get essential embedded Linux features incorporated into the mainstream kernel. For more almost 15 years, the various BSD variants have been (without compulsion) sharing their respective code, but still pursuing separate projects — a canonical example of the exaggerated fear of forking held by many GPLniks.

Sometimes forking is justifiable, if, for example different efforts have irreconcilably different goals. A better choice than forking would be a modular design that allows meeting goals via incremental improvements rather than re-implementation (a pet peeve of mine as a software engineer and manager for more than 25 years). In other cases, “duplicative” investment harnesses competition as a way to choose the best solution from a range of possibilities.

But in many cases, it’s merely pride — i.e. NIH (Not Invented Here) — that leads to forking. Fortunately for medical research, the other NIH (National Institutes of Health) directly or indirectly fund the bulk share of U.S. public research. So if the NIH says “no” to NIH — and yes to cumulative innovation — then there’s hope for at least one sector to channel resources towards areas that do some good.

Technorati Tags: , , , ,

No comments: