>Are we *seriously* going to end up with an sd<some-ungodly-number> some-
>where?  That'd take a serious number of wide controllers to handle; I don't
>know of a system that could handle that many; hopefully by the time one
>exists, we'll have thought of something better.

no, of course not.  not in the natural course of events.  of course,
there's *nothing anywhere* that stops anyone from putting

    sd24598 at scsibus0 target 0 lun 0

in their kernel config file.

some might argue that we should continue to support that, but i think
that's silly.  that would imply continuing support for all possible
uses and side effects of every facet of the system.  i, actually, had
a use for a security bug in something somewhere once, that i could use
to do a specific task.  then, when it was "fixed", my program stopped
working.

what might make more sense (visually, at least) is if (a) the use of
the upper bits was a separate set of device nodes (ie, sd0a-sd0p
instead of sd0i-sd0p) like freebsd and (b) ls printed the "middle"
bits separately as bsdi does it.

eg, you have two sd0a (and sd1a) devices: 4,0 and 4,1,0, but the first
set only goes up to sd0h (4,7) before switching over to sd1a (4,8)
whereas the upper set would continue up to sd0p (4,1,15) before
getting to sd1a (at 4,2,0).  the 1,0 or 2,0 would just come from a
visual split in the visual representation of the minor number.

am i making any sense?

-- 
|-----< "CODE WARRIOR" >-----|
codewarrior@daemon.org             * "ah!  i see you have the internet
twofsonet@graffiti.com (Andrew Brown)                that goes *ping*!"
andrew@crossbar.com       * "information is power -- share the wealth."