If someone purports to maintain an installation tool, then they should
test it. I still don't know what specific sorts of problems you're
talking about, that crop up only in an installation context. I don't
know of any, since PCMCIA is generally configured with the same tools
and scripts as in normal use.
> In contrast, the regular built-in device drivers almost never have any
> trouble at installation time, because they have been extremely heavily
> tested every single bootup to make sure that they always find the device
> reliably. They are known to not need any special parameters in 99% of all
> cases.
That's great. Unfortunately, things don't work like this in my world.
Maybe I should forward you a couple days' worth of PCMCIA problem
reports so you can get a feel for the sorts of problems I deal with
that don't fit this model.
> If you look at the PCMCIA stuff, because of how it is set up, special
> parameters are _accepted_ and considered to be perfectly fine. And that
> means that they tend to flourish, rather than be frowned upon. All of
> which means that installation is painful - once it is all installed it
> works beautifully.
I've invested a lot of effort in minimizing the need to customize
parameter files for individual machines, since this is responsible for
a large fraction of support issues. Believe me, every entry in the
HOWTO where I have to document a trick for some oddball laptop is
painful to me. But over time, the number of such tricks has gone
down, not up. Much of my effort has been mandated by the lack of
Plug&Play support in the Linux kernel, since most of these parameters
have to do with memory, IO, and interrupt resource tweaking.
I ported the Linux PCMCIA package to BeOS last summer, and guess what,
they generally don't need to tweak special parameter files. They've
got kernel Plug&Play support.
If you can suggest another way to do this, then I'm all ears. I think
we would both be better served if you spent a little more time asking
me, rather than telling me, what would make these things easier. I
think you're being incredibly arrogant.
-- Dave Hinds
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/