In message <199904071629.MAA01175@grosse.fugue.com>,
Ted Lemon writes:
>
>People used to express angst pretty frequently about how lame it is
>that NetBSD couldn't automatically detect the blank irq and I/O spaces
>that exist on some laptops - e.g., the NEC Versa 6030X.  The blank
>spaces exist because some irq lines are reserved for built-in devices
>that we usually disable.
>
>It turns out that Windows NT has the same problem.  I think the only
>reason Windows 95 seems immune to the problem is that vendors all ship
>custom versions for their laptops with these sorts of problems fixed.
>Windows NT 4.0 simply fails to use the 3c589 ethernet card on my
>laptop, and NT 4.0SP4 hangs solid if you try to use it.
>
>So buck up, dudes and dudettes - we're actually trying to solve the
>problem (or have we already solved it and I don't know?), and we know
>how to work around it, and our kernels don't crash as a result of it.
>We have no reason to beat ourselves up about this.

hi Ted,

The impression I have is that while what you say about Win95 and NT is
spot-on, Linux manages to get 'round this. And that's what we should
be aiming to compete with.

As best I uNderstand it (and I havent looked in years) the Linux stuff
wins party by being loaded as kernel modules, and partly by using a
smarter interrupt probe: before using an IRQs for PCMCIA, the Linux
drivers try and tickle a card into interrupting, and if it doesn't
interrupt on a given line, then that IRQ isn't used.

I've used David Hinds' Linux `pcmcia services' on all sorts of weird
hardware, for years (before NetBSD had pcmcia support) and I never
never saw any such troubles.

Does that fit with what you know?