On Wed, Sep 07, 2005 at 12:20:34AM -0700, W. Wayne Liauh wrote:
> Sven Luther wrote:
> 
> <Bah, most people keep proprietary code in linux kernel modules, and this is
> mostly ok and accepted. After all the kernel-module interface provides an as
> good code separation as the kenrel-userland interface does, at least if said
> modules do not use GPLed callbacks in the kernel, but they usually don't need
> to.>
> 
> The problem with using loadable modules is that, as someone in this forum has 
> educated me, Linux kernel developers have a strong prejudice against APIs 
> needed by proprietary loadable modules.

Well, it is a well defined API for third party modules. and you are supposed
to follow the rules, or expect random breakage all over. This is the same on
any kernel out there (solaris, microsoft, whatever) and a rule driver writers
have to conform too. 

The only difference with the window situation is that here you have the full
code, so it is more tempting to access any random bit of kernel code, while
you have to do some positive reverse-engineering or have insider access in the
proprietary kernels.

This is not a licencing problem, but a discipline and conformance to API
problem. I suppose if driver writers are unhappy with some of the API, they
can voice their opinion on LKML, and request a modification of the module
visible API, the same as if someone was writing drivers for proprietary
kernels, could ask for an API modification.

I leave it to you to bet on where such an API modification is more probable,
with the linux development model, or for windows drivers for example.

It is still a licence problem though if i write some kernel code, and declare
that i don't want it ot be linked with non-GPL compatible modules, then this
code cannot be part of the well defined module API, obviously. It could be
replaced if needed though.

Friendly,

Sven Luther

_______________________________________________
opensolaris-discuss mailing list
[email protected]

Reply via email to