Re: The LCC is a bad idea, but that doesn't mean the LSB doesn't have any issues
Op do, 16-12-2004 te 14:38 -0800, schreef Bruce Perens:
> Wouter Verhelst wrote:
> > To address these issues, the Free Software people created the LSB
> When I founded the LSB, the job I proposed for it was to do what the
> LCC is now proposing to do. I didn't believe that a paper standard
> alone would be effective at resolving cross-distribution compatibility
> issues, and it has proven not to be. A paper standard is certainly a
> helpful thing to have around, though, because it helps focus policy
> discussion and record the result.
> > That is, indeed, one way to solve it: throwing out the Freedom we, Free
> > Software developers, have, and slowly starting to move towards a
> > non-free platform.
> This is sophistry. LCC would be constructed of Free Software and would
> itself be no less free than Debian main.
I don't know what the essense of Free Software is to you; I suppose it
is something different to each and everyone of us.
However, to me, the essense of Free Software is that it allows one to
modify the software as one sees fit. Remove that ability, and I don't
see the software as Free anymore.
By requiring software to be bit-by-bit identical to a given binary
compilation, we would (at least partially) lose the ability to modify
the software as we see fit, and thus, the software wouldn't be as Free
This would be just a minor problem to us Distributors (who get to have a
say in the LCC); however, it would be a significant problem to people
interested in using now LSB, then LCC-certified software on platforms
with extra patches (say, they recompile the kernel to include a specific
feature that wasn't in the official LCC-certified kernel); they are at
risk of losing any and all support they get from the non-free developer
because of doing so.
This must not happen.
> > They should tell us where the standard is not clear enough, or where the issues with the LSB are, so that they can be solved.
> Here I can turn your own argument upon itself. Why do we create paper
> standards? One important reason is that in the world of proprietary
> software, vendors would not share their source code, so they had to
> define what the program would do in human language instead.
> A paper standard is a second-hand description of something that our
> community has the unique ability to share, bypassing the paper.
That is simply not true. Our community has the ability to share source;
that does not mean we should start sharing binaries, too.
> The essential technical problem with the LSB is that it is not
> sufficiently descriptive of every thing that a modern application must
> depend on,
I acknowledged that.
My point is that one can, indeed, solve this by providing common
binaries, but that doing so is the most ugly way to attack the issue.
If the paper standard is insufficient, it should be improved. If there
are issues with the toolchain that reduce the portability of the
produced binaries, then the toolchain should be improved. If the
non-free software wants to do things that are by its very nature not
portable, well, then it should not do that.
I do, however, not accept that the only way to improve the portability
of non-free software would be to make Free Software less Free.
> and is not within 10 years of being sufficiently descriptive at the
> rate that its development has been going. In addition, it is not the
> best possible technical solution to getting a bunch of Free
> Software-based distributions to be more compatibile with each other.
Then we will need to improve the compatibility in more ways than by just
providing a paper standard. There are many technically elegant ways to
do that. Using the same binary on all distributions is /not/ one of
smog | bricks
AIR -- mud -- FIRE
soda water | tequila
-- with thanks to fortune