[lkml]   [1999]   [Dec]   [6]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: Binary drivers
Date (Kendall Bennett) writes:

> Stephen Frost <> wrote:
> > Come up w/ the API. Develop the code. Modify the current driver
> > set (completely) to this new API. bug-proof it, test it, and prove
> > that performance doesn't decrease any for any of those drivers.
> The problem is a philosophical one. I am not willing to expend any
> energy helping to develop code to make Linux better if the end result
> is that it won't be accepted into the Linux kernel.

Actually there is no reason it needs to be accepted into the linux
kernel. You can just use it as a loadable glue module (I believe the OSS
guys use a similar scheme). You just have to maintain the glue module
(or at least keep it free so that others can fix it).

This solves the API problem, although not the random bugs in drivers
problem. Unfortunately Linux will be blamed if the binary only drivers
crash :-/ I fear encouragement of binary drivers will lead to a true
"linux headed for desaster" on a longer time.

Microsoft has a similar problem BTW, they say that a significant part
of the stability problems in NT4 came from third party binary only drivers,
and having seen the code of some of the vendor drivers recently released
for Linux I have no problems to believe that. I don't say that your
drivers are bad [I don't know], but you will certainly agree that there
are a lot of bad drivers around.

Your example of Solaris being more stable with binary only drivers
was a bit bogus BTW, because a Sparc usually runs with Sun provided hardware
and drivers [so Sun can fix them]. MacOS I would not cite as an example
for stability.Netware neither [Netware with random third party NLMs is hell].

> At the end of the day, maybe one day I will end up doing this, and
> the result may well be a fork of the Linux kernel (since there are a
> lot of people who do agree with what I have to say).

As I said, there is no reason to keep such a layer in the core kernel,
it can be a loadable module developed outside the kernel tree without problems.


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

 \ /
  Last update: 2005-03-22 13:55    [W:0.072 / U:2.608 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site