PERFORCE change 38917 for review

[Poul-Henning Kamp]

In message <20031001142349.A87379 at FreeBSD.org>, Juli Mallett writes:
>* "M. Warner Losh" <imp at bsdimp.com> [ Date: 2003-09-30 ]
>	[ w.r.t. Re: PERFORCE change 38917 for review ]
>> In message: <Pine.BSF.4.21.0309301448160.53279-100000 at InterJet.elischer.org>
>>             Julian Elischer <julian at elischer.org> writes:
>> : Someone who understands all this stuff should update
>> : 
>> : /usr/share/examples/drivers/make_device_driver.sh
>> 
>> That's premature at this time.  We're trying to hash out the
>> guarnatees that various subsystems need to make in order to ensure
>> that we can successfully detach drivers.
>
>In other words, the strawman is being used to find the fundamental
>flaws/races/... that we need to grow the APIs/etc. to accomodate?

Yes, that's the concept.

It is rather trivial to solve the actual problems, throw enough mutexes
at it and you're done.

It is far more tricky to do it in a way where you get both readable
source code (so people understand what it does and how it does it),
simple conceptual models (so people can understand it) etc.

And then of course, once we have that we need a transistion plan,
and that in itself is a problem because we want to minimize code
sweeps, and we have a lot of practically untestable drivers.

>Groovy.  Better APIs and primitives :D

Yeah, well, eventually.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.