John L. Ries wrote:
On Fri, 16 Apr 2004, Waldek Hebisch wrote:
Frank Heckenbach wrote:
Since there was no reply so far: IANAL, but AFAICS, the main question is whether unreleased software (which is a derivative of GPL work, but wasn't put under the GPL) can/should be called free software, or whether this question is relevant at all (-; since the only one who has to the software at all also has the freedoms).
Anyway, Peter Gerwinski (who originally wrote the above paragraph) suggests the following clarification:
[...] If you choose to release it, it must be under the GNU GPL. or
[...] It is not necessary to release it, but if you do so, it must be under the GNU GPL.
IMHO the clarification is less accurate then the original. GPL requires derivative works to be free software, but one can choose different licence for new parts (say MIT licence). In fact, one can invent "Fixepoint PublicLicence" (FPL in short) which is a copy of GPL, but requres that all derivative works are (if) distributed under FGPL. AFAIKS FPL is incompatible with GPL, yet a lot of folks thinks that GPL is FPL.
The GPL *does not* require derivative works to be free software *unless* they are distributed and even then, the copyright holder can make exceptions. I could, in theory, take GPC, add my own secret features to it, use it to compile my own proprietory software, which I could subsequently license for profit, all without violating the GPL, as long as I don't distribute my modified GPC under any terms other than those provided in the GPL; RMS might not personally approve of my conduct, but I would be within my legal rights.
The example is, of course, entirely hypothetical.
For real example look at Stanford Checker: http://metacomp.stanford.edu/
However, my point is that the words "free software" are fuzzy, and one may still claim that your modified GPC is free software. On the other hand fraze "must be under the GNU GPL" is _very_ precise and false. The proposed update would replace something that just stretches the truth by something that is plainy false.