Reasoning behind the "preferred form" language in the GPL

i think the crux of the matter is that when the GPL grants one the right to modify the distributed work, it doesn't say anything about what those modifications may be (as in, what kind of modifications the distributed 'source code' must allow). and clearly, there's a disconnect between what the RHEL6 srpm allows vs. what many people would like to be able to do as 'modification' (and what earlier RHEL kernel srpms allowed). so as you can read it in the followup post from Bradley Kuhn, the language of the GPLv2 is not what it should be, but that's too late now of course. that also means that my one-liner files would satisfy his 'tar x; make -C' criterion and therefore be compliant (and of course many other forms would also be compliant). sad or not, that's what we have to live with.

one last comment, as you seem to keep misreading the GPL: 'source code' != 'complete source code'. the sentence using the second term does *not* define 'source code', it defines 'complete source code' using the definition of 'source code' given in the previous sentence and i didn't mention it because it's irrevelant for the discussion: before we can talk about 'complete source code' we have to know what 'source code' is. apparently anything that compiles and a human can read in a normal text editor is good enough for 'source code'.