In message <20060220120240.GB8924@drowsy.duskware.de>, Martin Husemann writes:
>On Mon, Feb 20, 2006 at 05:53:20AM -0600, Peter Seebach wrote:
>>  The question of whether there's a higher-level
>> dependency is, it seems to me, specific to a given CONFIGURATION.  Which is
>> what config files are supposed to represent, no?
>
>IMHO the user should not be able to break this dependency by editing a conf
>file. The dependency is inherent in your code, nothing the user could select
>or where he has any option.
>
>But this is a grey area, I admit.

But which code?  It's not inherent in the PCI bus, or the processor, or even
any particular piece of hardware.

I guess you could make a case for expressing it as part of machdep.c, but even
then it's not quite a good match.

And actually, thinking about it more, there could be systems on which there's
more than one possible way this could be done, and you'd want to indicate
which one.  It really feels like a configuration-level question, rather than
a code question.

This isn't an attribute of the emac driver, the PPC405, the seeprom driver, or
a generic PCI bus, so none of the driver code should be doing it.

-s