Re: Binary modules for 2.6

[Jan-Benedict Glaw <jbglaw@lug-owl.de>]

[-- Attachment #1: Type: text/plain, Size: 1845 bytes --]

On Wed, 2003-09-03 05:40:40 +0200, Dumitru Stama <dics@ecommerce.com>
wrote in message <123-216863213.20030903054040@ecommerce.com>:
> With the current layout of the kernel modules there will be no way of
> distributing binary kernel modules anymore. Considering the structures

No? If it were like this, I'd say I'd really like that:)

> that combine to describe the way module works and the alignement of
> those depending on the processor type even if that processor is i386
> compatible. Personally i think this is a good move for open source

Well, Linux aims towards source compatibility. If you want to ship
binary modules, you've got to do *lots* of compilation.

You've to compile for all and any large distributions, and additionally,
some people will ask you to compile with their very personal kernel tree
(thay might have added loads of patches) and their .config file.

> community but what are we gona do with the proprietary drivers that do
> not have the sources available ?

If it's like that, vendors of binary modules may face a lot of
recompilation then...

However, such things are usually done in another way. If you encounter
such situations, you'll write a binrary-only core module plus some glue
code which will compiled on the client's system. You hardcode your
binary interface, the customer compiles Linux' source-comparible
interface. That'll resolve the problem.

However, you see that it's a PITA to live with binary-only modules. It's
easier to release sources:)

MfG, JBG

-- 
   Jan-Benedict Glaw       jbglaw@lug-owl.de    . +49-172-7608481
   "Eine Freie Meinung in  einem Freien Kopf    | Gegen Zensur | Gegen Krieg
    fuer einen Freien Staat voll Freier Bürger" | im Internet! |   im Irak!
      ret = do_actions((curr | FREE_SPEECH) & ~(IRAQ_WAR_2 | DRM | TCPA));

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]