An explanation of the difficulty of solving licenseproliferation in one sentence

Brian Behlendorf brian at collab.net
Thu Mar 10 00:59:04 UTC 2005


On Wed, 9 Mar 2005, Forrest J. Cavalier III wrote:
> If you are referring to the OpenBSD refusal to accept the Apache 2 license
> was the market working as it should.  One or the other will come around
> eventually, or they won't.  (It isn't clear to me which the "right" side of
> that particular case is, but OpenBSD was full aware of the consequences
> of going it alone.  I wonder if ASF was.....)

Theo's reason for refusal was:

"We've been clear: Their new license contains more stuff, and we do not 
accept MORE STUFF in licenses." - Theo

During the 2.0 process (which was more than a few years, and a very open 
process) we added requirements with *extreme* reluctance.  The one that 
has been most controversial was the patent defense clause - which by 
itself is almost so toothless as to almost not have been worth it, because 
it only harms a company using a Apache package that also wishes to sue 
someone over patents in that same Apache package.  Most likely someone 
wishing to pursue a case does so because they have a competing package the 
Apache package disrupts, so they are unlikely to be major users of the 
Apache package anyways.  We could have made the language more severe - 
applying to any Apache package the litigant uses, or even any Open Source 
package the litigant uses.  I think we could have done more, and Open 
Source projects would be wise to consider similar language.

OpenBSD made their decision after our multi-year process concluded, and we 
didn't discuss the possibility of a move like that, but I can see they 
have no choice if they wish the whole of OpenBSD to remain under the BSD 
license and only that.  Apache for OpenBSD is still available as a port; 
this was for the version included by default as part of the core OS.  For 
which they probably wanted a much more stripped down web server anyways.
Anyways, yes, it is an example of the marketplace working.  It's not 
really affected Apache anyways, as I can't say we've had much contribution 
from OpenBSD prior to the license change anyway.

I'm sympathetic to the passion in Theo's words - Apache's license is so 
permissive for the same reasons, we want our works to be the "Intel 
Inside" of other people's works no matter what license, and believe in the 
power of the carrot rather than the stick.  But I also believe that the 
real clear and present danger that Martin Fink and others should be making 
noise about is the potential for patents to cause massive pain for all of 
us.  An out-of-sight, out-of-mind approach to patent licensing like the 
MIT and BSD licenses take isn't enough anymore.  The problem is so 
important, it may even be worth an additional OSD clause requiring some 
mutual defense in new licenses to be approved.

The patent defense language the ASL 2.0 adopted happens to be incompatible 
with the GPL, despite involving Eben in the drafting of the ASL 2.0 and 
not hearing about the incompatibility until after the release.  It's our 
desire and intent that GPL projects be able to incorporate Apache-licensed 
works into GPL projects, despite the fact that we can't do the opposite. 
We're just nice like that.  So this is a defect; we're working with Eben 
now on how to fix it, either in a GPL v3 or an ASL 2.1.  If we do a 2.1 
the intent would be to license all future Apache releases under that 
license.  We don't have a relicensing problem because we've taken care to 
obtain permission from contributors to release code under the license of 
our choice.

 	Brian



More information about the License-discuss mailing list