On Fri, Jun 13, 2003 at 12:57:13AM -0400, Greg Pomerantz wrote: > Branden Robinson <branden@debian.org> wrote: > > I would say that the controlling preference is that of the person who > > last modified the Work and distributed it in that modified form. Anyone > > downstream from that person would have to keep the "source" in that form > > and the "binary" together. > I think one formulation that makes this a bit more explicit is this: > 4) The freedom to change the Work for any purpose, to distribute one's > changes, and to distribute the Work in modified form. Unrestricted > access to the form of the work which is preferred by its author for > making modifications, if applicable, is a precondition for this. > (with the possible addition of the words "or translator" after > "author"). This makes explicit the fact that there is no single > preferred form. If you allow individual authors to define their own > preferred version, you solve problems like this -- In trying to specify, this phrasing overlooks other important "preferred form"s. Someone may make changes to an existing piece of software which involve rearranging the APIs. Is this a "translation"? I'm not sure it is; but it may change the source sufficiently that the original author would prefer not to use that form as a basis for further modifications. The important nuance here is that the "preferred form" is the form that the last person to modify the code worked in when making those changes. This is unfortunately hard to specify precisely in a definition. > Andrea writes a program in C. Marty rewrites Andrea's program in > Fortran. Ulysses gets Marty's binary and asks for source -- Marty > can send his Fortran source because it is the form Marty prefers for > making modifications. > Alternately, assume Marty lives on a "proprietary island" and only > has access to proprietary programming languages. Marty should not be > barred from translating Andrea's program into a proprietary language and > distributing his modifications in that language (which is "preferred" > because its the only thing he's got). > I think any definition of preferred form needs to pass this test. In > other words, I think that any definition of "preferred form" that > requires an "open" or "transparent" format will be non-free. The same > holds true for document formats of course. The person who aims to > prepare a derivative work should have the option of using whatever form > she prefers, and should have no obligation beyond the distribution of > modifications in her preferred form. I am not sure at this point the > extent to which certain exceptions need to be in place in the case > where the author has some vested interest in selling you a proprietary > interpreter (in the extreme case, "pay me $100 for the AES key you need > to decrypt my preferred form"). Any thoughts? I think there are two distinct issues here that need to be considered: whether a piece of software can be considered free if the only source available for it requires proprietary software to be useful, and whether a license can be considered free if it enforces redistribution of source in a form that is useful without proprietary tools. The first question seems to be the more important one to this discussion, since being able to use/compile/edit the software is more fundamental than being able to redistribute it in modified form. And I think the intuitive answer here is the correct one: for purposes of identifying whether a piece of software comes with enough source code to be considered free, we already have a separation in Debian between the main archive and the contrib archive. Software in contrib is *effectively* non-free, even though it's not *intrinsically* non-free; it's non-free by circumstance. Free Software is well out of its bootstrapping infancy, and being self-hosting is an important characteristic of Debian. I think this formulation of freedom in response to the first question leads naturally to an answer to the second question: yes, we recognize certain copyleft compromises as being free, and one of these is that a license can require the redistribution of source code under free terms. If source code written to a non-free interpreter is effectively non-free, then I think a license which doesn't recognize distribution of this sort of source code as fulfilling the source distribution requirements *is* free. I'm not sure such a license requirement is necessarily a good idea, and I'm not sure whether the GPL contains this requirement, but I do think such a requirement would not render a license non-free. -- Steve Langasek postmodern programmer
Attachment:
pgpm87TIflahI.pgp
Description: PGP signature