On Sat, 3 Oct 1998, Greg A. Woods wrote:

> > 3cocm ards on PCI buses attach in bus and slot order.
> 
> If they can do it, why can't all PCI cards???

They all can. The biggest problem is usually determining the order,
since that's often not obvious from the motherboard layout. So what
do you do? Boot the kernel and read the messages! At which point
we're right back where we started.

> > Well, that would cover a hell of a lot of the machines we run on,
> > if not all of them. What's the point of implementing your scheme
> > if it's useless on most machines?
> 
> That's a good question, though I question your bias somewhat.  It may be
> based on counts of actual machines, but it seems not on counts of
> architectures.  Certainly the sparc and some m68k machines (incl. sun3)
> are more predictable, as is the vax and the pmax to the best of my
> memory.

Hello? Have you looked at what's in Sparcs these days? PCI slots.

> Certainly what I'm suggesting makes more sense on big servers which
> might actually make use of it than it does on wee personal machines.

Actually, it's the big servers that are really loaded with PCI buses.

> However if I've got a big Alphaserver with a half dozen
> SCSI host adapters on it, then I'm really going to appreciate the
> clarity of a descriptive system.

The Alphaservers in particular. :-)

> I dunno.  Does NetBSD want to attract the LISA crowd who are doing big
> 24x7 operations?

You mean the LISA crowd that are perfectly able to compile their
own kernels to hardwire devices however they wish?

> Even on the machines where there are turbo-channels with EISA or PCI
> busses sub-attached?

Digital machines these days don't have turbo-channel busses; they're
entirely PCI based, and have been for a long time.

cjs
--
Curt Sampson  <cjs@portal.ca>  604-257-9400    De gustibus, aut bene aut nihil.
Any opinions expressed are mine and mine alone.
The most widely ported operating system in the world: http://www.netbsd.org