Hi,
I was at Linuxtag and held up the virtual flag of Linux USB.
First of all, there were women in business attire. I kept feeling
burgeois.
I was asked what was the most important change in 2.5 for USB.
That caught me unprepared as I was thinking about 2.7. So I answered
without much relection, that the generic driver model was the most important
new feature for USB in 2.5. If you disagree, please yell now.
Now for users' (developers' really) demands:
1. Libusb is considered to suck. Users want non blocking IO.
2. Users want another private pointer in struct urb.
OK, now for my own thoughts on this.
The idea that usbfs needs replacement is not new.
At Linuxtag I've often heared comparisons of USB with networks.
Upon reflection it seems to me that people have a point there.
If we want clean non blocking IO we need to represent individual endpoints.
On the other hand the administrative unit that should be exported is
the interface.
As for a second pointer, although I agreed at first, I've now changed
my mind. It seems to me that the common pattern of use of the context
pointer of struct urb is stupid. In almost all cases we point it at a device
descriptor. This is quite odd, as the URB as addressed at a specific device.
We have a claim/unclaim model for interfaces, so an interface for which
there are active URBs should always have one and exactly one descriptor
associated with it. We should use that information and let urb->context
point at the associated scsi request or skb.
Comments?
Regards
Oliver
-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 -
digital self defense, top technical experts, no vendor pitches,
unmatched networking opportunities. Visit www.blackhat.com
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
