[bip] changing the Pygr license?

[Andrew Dalke]

On Sep 21, 2007, at 7:26 PM, Brandon King wrote:
> In my opinion, python bioinformatics code needs to be usable to as  
> many users as possible if we are going to compete against other  
> languages. In this case I would recommend BSD, but if you want to  
> force patches to come back to pygr, LGPL would be ok (although you  
> will loose some potential users compared with a BSD license). GPL  
> seems too limiting in this case because people are likely to  
> integrate pygr into other useful packages... and since many people  
> don't want their code to be GPL'd, they will not use pygr in their  
> project.

I've an anecdote to relate.

OpenEye is a software company doing chemical informatics and  
molecular modeling.
They used to have a product called "OELib" which was distributed  
under the
GPL.  After some time they found problems with that licensing model.   
It wasn't
that other software companies were taking their code and reselling  
it.  It
was that their customers, who are for the large part pharmas and  
biotech companies,
were making internal changes and not sending patches back to them.

I don't recall the reasons, probably because the pharma people would
have had to convince the lawyers that it was okay to do so.

They switch to a proprietary license, and only received one minor  
complaint
from their customers.  The old OELib code, which was GPL, is now  
available
as OpenBabel, and OpenEye has a new code base called OEChem.

In chemical informatics, and I think for bioinformatics, there isn't  
much
in the way of redistribution of software.  Most people using  
*informatics
software, and tweaking such software, don't want to be redistributors.

If you want to force patches to come back to pygr, use a license which
requires that.  Otherwise it won't happen.

As for GPL/BSD, I'm with Brandon on this, BSD over GPL.  But with the
proviso that since most people don't redistribute software, the license
doesn't really matter to all but a vocal few.

See also this commentary, linked to by a recent post of Titus'
   http://producingoss.com/en/producingoss.html#license-quickstart
which draws pretty much the same conclusion about the merits of
one of those licenses vs. the other.

				Andrew
				dalke at dalkescientific.com