Re: Anton's amendment
On Fri, Feb 03, 2006 at 12:28:08PM +0000, Matthew Garrett wrote:
> Anton Zinoviev <anton@lml.bas.bg> wrote:
>
> > You are not allowed to distribute a patch against work A which turs it
> > into a patch against work B. You are not allowed to do this because
> > this patch would be based both on works A and B. This makes it to be
> > "work based on B" so you have to distribute it in the form
> > original_B+patch. We have here a circular deadlock.
>
> "Patch" doesn't have to mean unified diff format. It's practical to
> produce a file that describes a mechanical transformation of work B plus
> an insertion of lines from work A. That's not realistically a derived
> work of B.
It is very questionable whether DFSG prohibits licenses that require
unified diff format (I won't be surprised if we already have such
licenses). But let we assume for a moment that acording to DFSG the
license should permit patch system of any kind.
Then your mechanical procedure must is able to insert in arbitrary
file some lines from work A (if your procedure is unable to insert
lines from A into files that do not belong to B, then it would be work
based on B). Suppose that you use this prodedure. The result would
be a file that contains portions both from A and B. If we want to
call this work free acording to DFSG the user must have a way to edit
this file. Suppose now that the user takes some text that originates
from A and inserts into the code for a function from B or makes any
other modification in the resulting file. You should have a way to
update your mechanical prodedure in a way that would make these
modifications buildable and I don't think this is possible if your
prodedure is not work based both on A and B.
However now I see that I missed another more obvious problem. You
have to distribute the combined work in the form original_B+
+patch_file_for_B. Instead you are distributing in the form
original_B+build_system_for_patch_for_B.
Anton Zinoviev
Reply to: