In message <199912140329.MAA13548@axp46.aist-nara.ac.jp>,
Toru Nishimura writes:

>> And that should be a 3 for the ioasic, no?
>
>It was my mistake.  I think it's a good practice to fall back to
>serial console when the framebuffer device which was designated by
>PROM is not available to booting kernel.  

Sorry, but I disagree strongly. If the user has configured the PROM
for a graphics console, then they probably already have a keyboard
hooked up. We have no guarantee that they have a serial console or
appropriate cabling.  If we can support some other video card --
great!  Falling back to a serial console, if a graphics console was
configured and NetBSD can find a graphics device it can support, is
just wrong. At least that's how it seems to me.


My recollection is that the DEC documentation suggested that OSes
should search for a console they can support, in the same search order
as the PROM does. (The documents were probably describing what Ultrix
did, but thats another detail).

> Extending the meaning of
>'console' variable for NetBSD local rule would be another story.

Yes, I agree, but that's not the case I'm talking about.