Alexandre Oliva

Alexandre Oliva at

the examples that first come to mind of non-equivalence and non-containment are:

* GNU autoconf; the license encompasses an additional permission that must be dropped if certain modifications are made.  as a result, in some cases you cannot distribute modified versions under the same license you could receive the original program.  this requirement in the complete licensing conditions appears to conflict with one of the bullets of the OSD, but it doesn't render the software non-Free

* Tivoized software: the OSI directors, back when I consulted them about this, didn't think that software distributed under tivoization-like anti-modification technical constraints caused it to fail to meet the OSD for the recipient, since they got the sources in conditions that meet the OSD (and the FSD), though the executable, in the tivoized scenario, is definitely non-Free Software

though these are real, rather than purely academic corner cases, and they could apply to an arbitrarily large set of software one way or another, nobody seems to be too worried about the differences.  those who care about OSS don't seem to care that autoconf-like licensing isn't strictly compliant with the definition, since they fall back to the FSD anyway, and those who care more about FS worry about tivoized software, but apparently don't assign so much relevance to OSS or the OSD.  so in the end the FSD seems to serve as the ultimate definition for everyone, though those who accept tivoized software also accept other proprietary software
GNU Autoconf is just GPL (a combination of GPL-2.0 and GPL-3.0, it looks like -- probably on a file-by-file basis for historical reasons).  There is a v3 exception that allows certain outputs (derivatives, as far as copyright law is concerned) to be distributed under other licenses, including proprietary licenses, but that doesn't make GNU Autoconf itself either non-free or non-open-source.  It just weakens its copyleft-ness a bit, but copyleft-ness is not required for FOSS.

Regarding tivo-ization, that's not an FSD-vs-OSD difference and it doesn't affect the covered licenses.  It's just a difference of interpretation, one that could be applied equally well to either the FSD or OSD.  The OSI's answer to the second point has always been that it's the license that's applied *to the source code* that matters.  In other words, it's not even meaningful to say that a binary is "free software" or "open source" in the absence of source code anyway, and if you do have the source code than *that's* the thing whose distribution under the license matters.

For the record, this is what happens every time I try to get a concrete example of any actual difference between "free software" and "open source software": it turns out it's either not a difference at all or it's a difference of interpretation that depends on the party doing the interpretation (i.e., not on the FSD or OSD itself).  In other words, the FSF interpreting the OSD would come to the same conclusion (in these situations) that they come to with the FSD, and the OSI interpreting the FSD would come to the same conclusion they come to with the OSD.

In more than 20 years, no has has yet shown me a license whose terms clearly match one definition but not the other.  I am fairly sure that no such license *can* logically exist, even in principle, and my attempts to find a disproof of this belief have thus far failed.

Karl Fogel at 2020-12-28T20:55:26Z

Well, I should correct one thing:

There may be instances where the FSF would come to a different conclusion, regarding a particular case, when interpreting the OSD instead of the FSD.  However, in the (rare) instances of this happening that I have observed, it has not been very convincing and I doubt a judge would agree with it :-).  That is, I think those differences of interpretation on the part of the FSF may be politically motivated rather than logically motivated.

Karl Fogel at 2020-12-28T21:07:00Z

> GNU Autoconf is just GPL

not true.  any exceptions *are* part of the licensing conditions.  that some part of the licensing conditions of a program is recognizable as a separate license doesn't make that the license of the program.  the license is the *complete* set of permissions and conditions, not a subset thereof.

> There is a v3 exception

that is part of the licensing conditions, and that cannot be extended to certain modified versions, which conflicts with requirements of the OSD.

that the OSD only looks at the license of the sources is exactly another fundamental difference that leads to different outcomes when assessing whether a program meets the OSS or meets the FSD.  the latter looks at conditions imposed on users through technical means, in as much as they impose substantial limitations on the essential freedoms.

that the OSD attempted to establish objective conditions that model the FSD but failed to capture this is IMHO a bug in the OSD, that most people of OSS allegiance weasel out of by dismissing it as irrelevant, even when they admit the different would stand in the way of the so-called open source effect.  double-thinking at its best ;-)

Alexandre Oliva at 2020-12-28T21:28:20Z