[License-discuss] dual licensing and the Open Source Definition

John Cowan cowan at mercury.ccil.org
Fri Dec 13 15:28:44 UTC 2013

zgilboa at culturestrings.org scripsit:

> One interesting side-effect of the above is that software can be
> released under a strong copyleft license, for instance the GPL,
> and yet be accompanied by the option to "buy one's way out of the
> license," thereby releasing the buyer from any and all obligation to
> make the modified source available to the public.

Technically, that is not correct.  Licensing applies on a per-copy
basis, so in that situation there are some copies of the source that
are available under an open source license and other copies that are
available under a proprietary license.  This is only possible for the
licensor, who is not obliged to follow the restrictions of his own

> In light of the above, and given the guarantee of the Open Source
> Definition with respect to source availability [...]

OSD #2 is narrower than people tend to think: it requires the licensor
to make source available, and to allow people who have that source to
redistribute it.

OSD #3 requires only that the license permit modifications and
derivative works: it does not require those derivative works to be Open
Source: see below.

> [T]o what extent does the GPL meet the OSI promise regarding the
> source of Open Source Software remaining open?

1) Entirely; and 2) There is no such promise.

The GPL applies only to copies of works to which it is applied by the
licensor.  If other copies exist under other licenses, that has nothing
to do with the validity of the GPL.

But more importantly, the fact that a work is under an OSI license does
not guarantee that all derivative works must be under OSI licenses.
Many widely used Open Source licenses such as the BSD/MIT family and the
MPL family allow for proprietary derivatives of open-source works.

> After all, if vendors can take GPL'ed software and buy their way out
> of the license so that binaries, with or without changes, can be
> distributed without restriction and without a corresponding source,
> then something is probably not working the way it was originally
> intended.

RMS probably doesn't like it, no.  But people can use the GPL however
they want to.

> 2) Consider the case of an individual entrepreneur who created a
> software library, and who would like to require vendors of commercial
> products that _depend_ on that library to pay a _one-time fee_, but
> otherwise be permitted to use the library or distribute it in any
> way they see fit without additional charges, and provided that the
> original source code, along with all changes that were applied to it,
> remain available to the public.  Would that author be able to release
> his/her library under an OSI-approved license?

In a word, no.  Requiring some users to pay for source and not others
flatly contravenes OSD #5.  But that's an ideal scenario for a GPL +
proprietary dual license.  Vendors of proprietary software who want to
use the library in ways the GPL forbids have to make terms with the
author.  What those terms are, whether a single payment or otherwise,
are entirely between the author and the customers, and no concern of
this list.

> Then again, it seems to me that the possibility to regulate one-time
> charges for commercial use from _within_ a license should be much
> preferred over a de facto option to bypass the license altogether.

It comes to the same thing: bypassing the GPL is bypassing the GPL.  And
since the GPL's author has forbidden people to make modified versions
of it, alternative terms must be placed outside it.  This makes sense
anyway in terms of the model in which some copies are available under
GPL and others are available under the proprietary license.

My confusion is rapidly waxing          John Cowan
For XML Schema's too taxing:            cowan at ccil.org
    I'd use DTDs                        http://www.ccil.org/~cowan
    If they had local trees --
I think I best switch to RELAX NG.

More information about the License-discuss mailing list