For Approval: Transitive Grace Period Public Licence, v1.0

[zooko]

[adding To: license-discuss; I'm not sure about conventions on these  
lists, but I would assume that this sub-thread is more germane to  
license-discuss than to license-review, in which case further follow- 
ups should probably be directed only to license-discuss.]

Dear Larry Rosen:

Thank you for describing the Aladdin Free Public License and the idea  
of "eventual source".  Note that this idea is different from the  
TGPPL in that the TGPPL is transitive (like the GPL and the OSL are),  
but the "eventual source" strategy of the Aladdin Free Public License  
was a one-shot deal from the original creators to original  
recipients.  This difference is crucial, because in the non- 
transitive case all that we are talking about is a copyright holder  
choosing to release a work under a proprietary licence, and then  
later choosing, for whatever reason, to release the same work under  
an open source licence.  With the TGPPL, what we are talking about is  
a recipient of a work choosing to create a derived work from it and  
redistribute his derived work *on the condition* that he will  
eventually (within twelve months) open the derived work.

This is a different situation from the "eventual source" case,  
because the actor is choosing whether to gain the advantage of re- 
using someone else's work, on the condition of eventually  
contributing to the world's library of open source software.  No such  
decision enters into the "eventual source" case.

However, even though these are very different legal and economic  
situations, they're obviously related, and I've tried to learn what I  
could about how "eventual source" has worked in practice.

As I've previously mentioned, there was an interview with L. Peter  
Deutsch by Stig Hackvan conducted in in 1997.  It is well worth  
reading!  (A-ha -- and on this re-reading of it I notice that he says  
Russ Nelson was considering using the AFPL.  :-))

http://devlinux.org/deutsch-interview.html

Also I inquired with my friend Raph Levien, who was partially  
responsible for Ghostscript development and licensing at one point,  
possibly even at the point that they dropped the "eventual source"  
idea and went to straight GPL.

He replied that the "eventual source" idea didn't work out that well,  
and that things were much easier once they switched to GPL across the  
board.  The major reason, he said, was that open source contributors  
were picky about not contributing to the AFPL branch, only to the GPL  
branch.  (Note that this is inconsistent with what LPD said in the  
interview, but Raph's involvement began approximately five years  
after that interview.)

Raph said that the GPL branch didn't compete directly with the  
licensing revenue from the AFPL branch (due to different use cases/ 
customers), which was fortunate, but which also meant that the  
Artifex company and the open source community weren't contributing to  
each other's needs as much.

As another data point, Id Software (makers of Doom and Quake) used to  
release its #N-2 game under the GPL at the same time as it released  
its #N game as a commercial product.  I don't know if they still do  
that, nor even if they still make games.

As another data point, I recently had a pleasant chat with Peter  
Braam, who was founder and leader of Cluster File Systems, Inc.,  
makers of the Lustre filesystem, during the entire period of its  
independent existence before it was bought by Sun in 2007.  During  
this period, CFS sold the latest version of Lustre to customers under  
the GPL, while releasing the 12-month-older released to the public  
under the GPL.  (Yes, you read that right. Here is an interview with  
then-CEO Phil Schwan which mentions that licensing strategy: [1] .)

Peter Braam told me that the customers didn't like the existence of  
the commercial version and repeatedly insisted that the public should  
be given access to the most recent release.  However, none of them  
chose to exercise their GPL rights to redistribute the commercial  
release to the public.  Due to this pressure he began releasing  
Lustre to the public earlier, which he says greatly *increased*  
revenue and "improved relationships with customers, but also led to a  
considerable increase in non-paying users".

Peter also stated that there was no way Lustre could ever have been  
developed, at great monetary cost, if those same customers (almost  
all from the government defense community) hadn't been paying to get  
Lustre developed.  He said that if he had to do it again, he would  
need serious funding from someone, and that if that funding would  
come from investors, he thinks a plain proprietary approach might be  
helpful.

In 1997 Sun bought Lustre, and announced [2] that they would be  
releasing everything to the public under the GPL and selling support,  
which have they have proceeded to do.  If anyone has numbers about  
Sun's profit from that strategy I would like to learn.

I have a final data point: Tahoe.  It has been distributed under your  
choice of GPL or TGPPL since its inception.  (Technically, a  
predecessor to TGPPL v1.0 for the first few releases.)  The Free/Open  
community seems happy to contribute to Tahoe so far.  It's a young,  
small project, but we're having a lot of fun with it.  Here is the  
CREDITS file which lists all of the people who've contributed  
something to Tahoe: [3].

By the way, my motivation in writing about these empirical  
observations is not so much to persuade you to use TGPPL; it is more  
just to state what I know for everyone's reference.

Regards.

Zooko Wilcox-O'Hearn

[1] http://www.wwpi.com/archive2/300-12-questions-with-phil-schwan
[2] http://www.sun.com/software/clusterfs/general_faq.jsp
[3] http://allmydata.org/trac/tahoe/browser/CREDITS
---
Tahoe, the Least-Authority Filesystem -- http://allmydata.org
store your data: $10/month -- http://allmydata.com/?tracking=zsig