Re: [9fans] TODO lists for Plan 9

[lucio@proxima.alt.za]

> I'm not sure that a mini-OS makes sense; I'd rather just see the
> standard USB support improve.  A good first step would be to gather
> together all the available USB documentation, starting with the actual
> on-the-bus protocols and programming the host interfaces (OHCI, UHCI,
> etc.).  That will be a fearsome pile, alas.

I have such a thing, the paper version is not even luggable.

> The first obstacle to using USB for storage devices is that USB 1 is
> slow and USB 2 (the original USB 2, not the USB 1 that has been
> renamed to USB 2) is fast enough but requires new code to run OHCI
> interfaces (yes, that's redundant).

Huh?!  You are not serious, are you?

I was under the impression that matching the 480Mhz bus speed of 2.1
would be hard to do, what with the need to poll devices.  But I am not
speaking form a position of knowledge, most of what I know is very
superficial.

What I did figure out, is that 2.1 uses EHCI and falls back to OHCI
for the slower speeds.  It makes it look like OHCI suffices, but then
the speed is nowhere near as fast as the real thing.  Good enough for
a suprprising number of products, but one feels a little cheated.  I
know NetBSD is still developing EHCI but seem to have a working
version of OHCI (and UHCI, of course).  I don't know where Linux is
at, or Windows, for that matter.

BTW, the Bluetooth spec (for comparison purposes) is 1200 pages, which
is not trivial.  Fortunately, that includes details the average
developers won't need, but it is still a bit much, specially when my
immediate interest is to drive a USB/Bluetooth device :-(

As for the mini-OS, it is a very vague idea, but it caught my
imagination.  My foggy image is of a xen-like thing at the core of
things and the guest OSes knowing how to talk to each other in a 9P
fashion.  Is that totally absurd?

++L