A scientific basis for Open Source Software

Yup. Or just speak the university lawyers earlier in the process.

Shrug and say, "Look, this might be useful to some researcher somewhere, but it would be an epic failure as commercial technology transfer. It's just half-baked, and the entire market is probably a half-dozen other broke academics. If we sunk in $50–150K of startup costs and a year of our professional lives, we might scrape up $5K of sales if those academics busted their budgets. Frankly, we're better off spending the time writing more grants."

"If you want to maintain some commercial rights on the off chance that anybody, anywhere ever cares, we can go ahead a slap a strict 'share and share alike' license (such as the GPLv3) on it, and reserve the right to license it under alternate terms if somebody wants to pay us."

I've successfully made this pitch at two major research universities. In my experience, if you have the backing of your PI, your lawyers will probably go along.

As much as I dislike the GPL, it has an additional benefit in science: with some grants, you have to transfer copyrights to all software that is produced, *except* when you can't. In such cases, if you can argue that you need to expand an *existing* GPL-licensed project, you can often get a waiver (for copyright transfer) for deliverables that involve that particular software.

In other words: licensing scientific software under a strong copyleft license could make life easier for other scientists who want to expand your work.