Subject: Re: OSL for libraries (was Re: Submitted for Approval: OSL 3.0 and AFL 3.0)
From: "Gordon Kindlmann" <gk@bwh.harvard.edu>
Date: 12 Sep 2005 12:34:36 -0400

On Sep 12, 2005, at 11:13 AM, Lawrence Rosen wrote:

> Gordon Kindlmann wrote:
>
>> * Is the author of such a collective work required to notify
>> users that the collection includes a copy of my library, by
>> identifying the name of my library, or by duplicating the
>> copyright notice that accompanies my library?  This is
>> required by LGPL, FLTK, and FOX.
>>
>
> Copies of the Original Work (your library) must be made available  
> in source.
> A copy of your library is being distributed, remember.
>
>> * If there is such an attribution requirement, how can I waive it?
>> wxWindows does away with any attribution requirements.
>>
>
> A licensor can waive any provision of his license with respect to  
> his own
> original works.

I believe I'm now grasping the simplicity of OSL 3.0, which deftly  
leverages the existing legal framework for collective and derived works.

My concern, again from the standpoint of applying OSL 3.0 to a  
library, is that the spirit of things like LGPL (and the exception  
notices that have evolved on top of it) is that at least from a  
programmer's standpoint, there is a meaningful and fundamental  
difference between the distribution of a modification of a library,  
and the distribution of an executable which links against the  
unmodified library.  If my library is under OSL 3.0, both a modified  
version of the library, and a copy of the library as contained in an  
executable (aka collective work), are also to be licensed under OSL  
3.0.  Correct?

So, when you say that "a licensor can waive any provision of his  
license with respect to his own original works", is there any  
precedent for limiting that waiver to certain kinds of copies or  
derivatives?

In the particular context of the OSL 3.0, this would seem to require  
some complication of Section 1(a) and 1(c), in that I want OSL 3.0 to  
cover modifications of the library ("Derivative Works" of 1(c)), and  
to stand-alone copies of the library ("Original Work" of 1(c)), but  
*not* copies of the library that appear within an executable  
(collective works of 1(a)).  When part of a collective work, and when  
in binary form, I would want the copy of my library to be licensed  
under, for example, *AFL* 3.0 instead of OSL 3.0.  Is this practical  
within the context of OSL/AFL 3.0, or possible at all?

>> * If my library in source form is licensed under OSL 3.0,
>> doesn't its compilation (in the computer science sense) into
>> binary object code form create a derived work of my library ...
>
> You're not being stupid. I do not treat computer-style compilation  
> as a
> translation for purposes of derivative work analysis. My reading of  
> the
> guidelines at the Copyright Office informs me that they treat  
> object code
> compiled from source as merely another form of the same work. They  
> allow you
> to copyright trade secret software by submitting compiled object  
> code under
> a "rule of doubt" which can be resolved, if necessary, by reference  
> to the
> source code. They appear to treat it as two forms of the same work,  
> and so
> do I.

Fascinating.  Does this doom my effort to use a reciprocal license  
for my library, but have non-reciprocal terms for binary copies of  
the library appearing in a collective work?  I would not want, for  
example, someone to contrive a collective work around a copy of the  
*source* of my library, using the non-reciprocal license I permit for  
collective works, lift the library out of the collective work, and  
then modify and distribute it, again under the non-reciprocal license.

Gordon