«

In message <f8egke$p9r$>, RL wrote:

> The attitude that closed-source
> drivers are evil, and nothing is to be fully supported unless it is also
> GPL'd and in the main source tree is the problem. If companies could
> produce binary-only drivers, that worked reliably from one kernel
> release to the next, then support would be greatly improved IMO.

That would simply put you back in the Windows situation, with chronic driver
instability being responsible for so many crashes.

This sort of thing has been discussed before. Greg Kroah-Hartman likes to
mention the example of USB: the Windows USB driver APIs were reworked about
3 times (with Vista probably counting as a fourth). The ones in Linux were
also reworked 3 times. The difference is that Windows still has to carry
around the obsolete backward-compatible versions and support them forever,
because there are old, obsolete, binary-only drivers that continue to rely
on them. Linux doesn't need to carry around obsolete baggage in the kernel
this way: when the APIs were reworked, all the drivers could also be
reworked at the same time to get rid of the old calls.

So you see, the Linux kernel is actually less bloated than the Windows one.

Lawrence D'Oliveiro wrote:
> In message <f8egke$p9r$>, RL wrote:
>
>> The attitude that closed-source
>> drivers are evil, and nothing is to be fully supported unless it is also
>> GPL'd and in the main source tree is the problem. If companies could
>> produce binary-only drivers, that worked reliably from one kernel
>> release to the next, then support would be greatly improved IMO.
>
> That would simply put you back in the Windows situation, with chronic driver
> instability being responsible for so many crashes.
>

There is another aspect - security. Paranoid users (eg SIS, GCSB, etc)
would not want binary 'blobs' in kernels or associated modules intended
for top secret applications.

By the way I have no idea what these outfits actually use - this would
be classified information in itself.

In message <46aaf38a$>, peterwn wrote:

> Lawrence D'Oliveiro wrote:
>
>> In message <f8egke$p9r$>, RL wrote:
>>
>>> The attitude that closed-source
>>> drivers are evil, and nothing is to be fully supported unless it is also
>>> GPL'd and in the main source tree is the problem. If companies could
>>> produce binary-only drivers, that worked reliably from one kernel
>>> release to the next, then support would be greatly improved IMO.
>>
>> That would simply put you back in the Windows situation, with chronic
>> driver instability being responsible for so many crashes.
>
> There is another aspect - security. Paranoid users (eg SIS, GCSB, etc)
> would not want binary 'blobs' in kernels or associated modules intended
> for top secret applications.

Unfortunately, that kind of thing is already present, e.g. in the madwifi
drivers.

In message <f8ehih$qqj$>, I wrote:

> In message <f8egke$p9r$>, RL wrote:
>
>> The attitude that closed-source
>> drivers are evil, and nothing is to be fully supported unless it is also
>> GPL'd and in the main source tree is the problem. If companies could
>> produce binary-only drivers, that worked reliably from one kernel
>> release to the next, then support would be greatly improved IMO.
>
> That would simply put you back in the Windows situation, with chronic
> driver instability being responsible for so many crashes.

And also, for example, these sorts of issues
<http://theinquirer.net/default.aspx?article=41440>.

The only way to root out crap drivers is to expose the source code to
ruthless peer-review scrutiny.