Are implicit dual-licensing agreements inherently anti-open?

[Brian C]

Hi David,

David Barrett wrote:
> It seems to me the OVPL combines elements of BSD, GPL, and MPL -- all of
> which are open.  Thus I see strong precedent to argue that the OVPL
> itself is also open.
> 
> Specifically, the BSD license, as Stephane mentioned, allows Microsoft
> to "take" code into a closed, proprietary branch and not contribute
> anything back.  I think we can all agree that the BSD license meets the
> "open source" criteria, so this seems to be solid evidence that an open
> license *can* allow code to move into closed branches.  In the case of
> BSD, anyone has that right.
> 
> The GPL, which I think we can also agree is "open", is at the opposite
> end of the spectrum.  It explicitly disallows anyone to take code a
> closed branch -- initial developer or otherwise.
> 
> The OVPL license, however, sits somewhere in between.  It is like the
> GPL in that contributors cannot take code into a closed branch.  But
> it's like the BSD in that the initial developer can.
> 
> So the question is: is that still open?  Well, if the BSD is open, and
> if the GPL is open, I see little reason why the OVPL shouldn't also be
> open, because it's just combining behavior between the two.

The main reason that I would imagine people might object to this mere
combination analysis is that in this case the combination results in
asymmetrical rights for different individuals. The ID is granted rights
by the license that the rest of the licensees aren't provided. Now,
copyright law always grants the copyright holder rights that aren't
necessarily granted to licensees, but there's little OSI can do about
the law. However, OSI could decide that when licenses enshrine an
asymmetry that there's a problem. (I'm not advocating a position right
now. Just thinking.)

When it's the ID that's getting greater rights, there is some rationale
for that, as it then serves as an incentive to open previously closed
code. In such cases, maybe an asymmetrical license is consistent with
OSI's principles, and you mention the MPL as possible precedent for this.

But, in general, I can imagine such a license that would be
objectionable, say one that gave only U.S. citizens the right to make
proprietary versions or gave only men such a right, or etc (fill in an
arbitrary characteristic not related to the origin of the code.) These
would seem to pretty squarely violate OSD #5.

So perhaps the OVPL presents a question to OSI: Is a license that grants
greater rights to an initial developer than it grants to other licensees
consistent with OSI's principles, in particular, does it constitute
"discrimination against persons or groups"?

If OVPL is approved, then I suppose we'll definitively know the answer
to that question.

I suppose one argument against allowing it is that it seems motivated by
recognition of prior contributions of time and money to develop the
initial version, but over time, a subsequent developer may have
contributed far more time and money such that the end product isn't even
recognizable as the original, but nonetheless the ID would still be
granted greater rights. This seems to break the meritocratic ideal that
supported the distinction to begin with.

Another argument would be that open source doesn't care about merit
among licensees; we're not worried about "freeriders". On the contrary,
we flourish because of them. So, any license that tries to distinguish
among people based on the size or worth of their contribution to the
project is simply misguided. (Again, I'm not advocating. Just speculating.)

Brian

> The only reason I can see is that the OVPL distinguishes bewteen
> contributors and the initial developer, whereas neither BSD nor GPL
> does.  So in a sense, it's discriminatory.  However, there's precedent
> for this in the MPL, as it makes a special effort to define "initial
> developer" in 1.6 (what precisely it does with that distinction is the
> subject of debate, but there is no debating that the distinction is in
> fact made).
> 
> So it seems to me the OVPL is taking the code-closing options of BSD,
> the copyleft of the GPL, and the initial-developer distinction of the
> MPL, and rolling it into one.  All of the pieces from which the OVPL is
> composed are open, and thus to my view, this seems strong evidence that
> the OVPL itself is also open.
> 
> Where is this analysis off?
> 
> -david