Dan Streetman wrote a lot of stuff about interrupt transfer model... <big snip>
> So, if my opinion counts for anything, I would vote for (first) allow > INT URBs to be queued and (second) remove URB > auto-resubmission entirely. <big snip> I was too lazy to write up such a detailed discussion, but I agree 100% with Dan's arguments. Adding re-submission to existing drivers is something that makes it *more* obvious as to what is going in (IMHO); the current auto-resubmission mechanism fooled me in the beginning, as well as is very impractical for interrupt out. Interrupt out seems to be the stepchild transfer mode. However, it seems that there appear more and more devices which have a use for it, and going forward, it would be very helpful to keep things nice and clean. Also, having the same paradigm for interrupt in and out, and similarly, bulk in and out, is a good thing - it's just far less confusing this way for new driver writers. And if it counts anything (I know it *doesn't* :), Windows' USBD has a very similar mechanism. Bulk and interrupt transfers happen the same way (even using the same structure), and in and out directions are treated the same; queueing is possible for both cases, the host controller driver just picks the next transfer in the queue when the first completes. Just my $0.0001 ..tom _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
