[Coco] Thought

[Mark McDougall]

Eric J. Rothfus wrote:

> This is where I wonder.  Although the interface circuit would be easily
> able to issue the USB request within a usual seek time, what concerns me
> is the PC responding to the request quickly.  Granted, if a program was
> written that would stay resident (ie - not have to be swapped in) and if
> the USB was quiet I wouldn't worry about it.  But that adds complication 
> to the PC side.

Any idea what a typical seek/step time would be? Or typical/allowable 
latency on starting a sector read? Remember on the PC side of things, the 
USB I/O is interrupt-driven. It has to be fairly responsive to be able to 
handle the USB 'Interrupt' transfers which are designed for guaranteed 
bandwidth and constant (predicatable & timely) latencies, used for input 
devices for example.

I worked on a linux USB driver where we were abusing bulk transfers trying 
to get single-cycle turnarounds on small (<64 byte) packets. We were able to 
achieve reliable 1ms turnarounds most of the time.

Also, CoCo floppies are very small and could easily be cached even at the 
driver level on any modern intel box.

> Anyway, back to my "paying" job... But this is a fun distraction.

Ah, work is vastly over-rated! ;)

-- 
|              Mark McDougall                | "Electrical Engineers do it
| <http://members.optushome.com.au/msmcdoug> |   with less resistance!"