Subject: Re: USB feedback wanted
To: Chris G. Demetriou <cgd@pa.dec.com>
From: Todd Whitesel <toddpw@best.com>
List: tech-kern
Date: 07/08/1998 18:24:50
> However, even if it is fixed to allow things like string locators,
> topology is still a valid identification mechanism for some uses.

Speaking as a Luser for a few minutes:

I would be very happy if the kernel used topology but with a simple and
_documented_ algorithm, so that whenever I am in doubt I can just eyeball
my USB rats-nest and figure out what order everything has.

For example, two speakers on the same hub ought to always have the same
relative device ordering as their USB port ordering. Something that was
discussed earlier involved a topologically-determined 5-digit number for
each device; something like that could specify the ordering which would
be used when numbering devices 0, 1, 2, etc. The only real question in my
mind is whether we order things depth-first or breadth-first, and that may
well be a religious issue.

I can deal with plugging in auxiliary speakers "after" my two main speakers,
if that's what I want. I can even move the speakers to a totally different
hub -- as long as I preserve the same 'scan order' then the device node order
will be preserved, and I'm happy. USB lets me easily unplug existing devices,
so I can do whatever it takes to get the relative ordering that produces the
device node ordering that I want.

I do NOT want to rebuild kernels with magic strings in them for USB.

I DO want magic strings for SCSI, however, so my second HD is always sd1
and my Zip drive can have sd2 reserved for it when it's not hooked up.
(Or zd0, that'd be nice.)

I do NOT want zillions of device nodes in /dev -- the ptys are bad enough!

Todd Whitesel
toddpw @ best.com