Re: transformations of 'source code'
Branden Robinson said:
> On Fri, Mar 07, 2003 at 05:36:57PM -0500, Joe Moore wrote:
>> Nick Phillips said:
>> > I don't think that losslessness is the right criterion, rather
>> something connected to the meaning of the source and the
>> achievability of the source's object.
>> Can have useful source recovered from it, in a form that is
>> <something> for modification?
>> I can't think of the <something> to put there. It's not preferable,
>> it's not easy, it's not acceptable... Any thoughts?
> Are you thinking of "suitable"?
That's a very good word. "form that is suitable for modification". Thank
> But the GNU GPL's definition of source code is more than that; it's not
> just a "suitable" form for modification, it's the "preferred" form for
> modification. I think we should preserve that.
I think there are too many problems with "preferred".
The author of software A prefers to work in Visual IDE 2000 (a proprietary
GUI for application development). This has something that's not quite
exactly like Makefiles, which only work under Visual IDE 2000. However,
he can automatically generate working sources and Makefile(s) (and
Makefile.in) from his preferred .ide format, by clicking the "Publish"
button. Why would the .ide files be the "preferred form" if no Free
Software developer can read them?
>> This requirement, while totally inadequate from a legal perspective,
>> also explains the foreign language: Speakers of the foreign language
>> get useful source from the transformations, and a second translation
>> can get useful source back into the original language.
> I don't agree with this analysis at all. Translations from one natural
> language to another are very lossy things. Ever read Shakespeare
> without using the footnotes? How about Chaucer? Magnify that problem
> by ten.
I'm not talking about literature, but I do see your point. On the other
hand, (translated) Chaucer is still meaningful and pretty good literature
in both French and modern english. It may not be identical to the
original, but the better translations are closer than the bad ones.
Another thing would be source comments: There's been a lot of work with
i18n of executable messages, but that doesn't help foreign developers to
improve the source code. Translating the (english) //comments into (for
example) elbonian may not help the original author (and may make it
exceptionally hard for him to maintain the code a few years later), but it
shouldn't be forbidden by the license, when it's exactly the sort of thing
Free Software is supposed to promote.
If an elbonian programmer translates the author's comments and
documentation into elbonian for his local hacker's club, and they improve
the original software, are they obligated to have their improvements
translated back to english?