glib dependency for the X Server

[Russell Shaw]

Dave Airlie wrote:
>>If an improved X is released as a closed binary on some embedded system,
>>who cares? It is in the interest of the vendors doing that to put back
>>improvements into the public X source to minimize their patch maintenance
>>when X development starts diverging off into hyperspace needing a quad-CPU
>>graphic card to get useable performance.
> 
> I do, I would like to use that work in some of my projects perhaps, or
> I would like to avoid reinventing the wheel, reskinning the cat,
> refixing the bug in some c+t chipset etc.. it should be in the
> interests of those vendors to contribute back something after all they
> just got a whole lot of code from others... but that is an MIT vs GPL
> fight which will only make me produce the GPL X server I've long
> threatented :-)

I should clarify. I dislike closed-source graphics drivers for users
desktop systems, because the inability to poke at it affects me and
other developers directly. If the drivers are stuck in some embedded
system, it just doesn't matter.

If i had the time and got around to figuring out pc graphic-card
programming, interfacing, and kernel interaction, i'd think about
doing a new X server that is one tenth the size of the current one,
and without any Xt related stuff.

>>The purpose of (L)GPL is to give more incentive to programmers to
>>contribute back work knowing that it won't be improved and sold in
>>closed binaries. The nature of one global X instance on a pc that
>>everything needs to be compatible with, fulfills that goal.
>>
>>I don't care if any improvements i make get sold as improved closed
>>static binaries in an embedded system, because i can do the same myself
>>if i want to, which is more freedom than (L)GPL.
>>
>>The problem with (L)GPL is that it lacks specific clauses for embedded systems.
> 
> You aren't making any sense with your statements treating the GPL and
> LGPL as the same thing, they aren't, I know they both have problems in
> embedded systems, but that first paragraph is only for GPL, you can
> put LGPL code into closed systems, of course you can you just have to
> make sure someone can replace it....

Thats the point. Many embedded systems have no linker or operating
system, and can have everything in an eprom.

I'm sure that if the binaries generated by gcc inherited the L(GPL),
then gcc would have few if any back-ends for non-pc CPUs, because
there'd be no closed hardware using it, designed by users or companies
that need a competitive advantage.

It is the same with X. Even if the actual ports of X to an obscure hardware
platform aren't released, if it's a common thing to do, then the infrastructure
for easy porting to diverse hardware will find its way into the public X source.