On Wednesday 10 March 2010, Alan Cox wrote: > It sounds like a bug that someone needs to provide a test case and > explanation for. The underlying buffering logic appears solid up to about > 40MB/sec (beyond that the existing buffer settings and round trip time > for flow control would need addressing and possibly some data handling). > > I don't see any real value in a char abstraction for most of the cases > where we hit performance and other problems - a packet abstraction ought > to be both a lot faster and provide a more flexible interface for > stuff where usb frame boundaries matter. Worth clarifying. If u_char is just to provide a lower overhead byte stream abstraction (packet/frame boundaries never matter) than TTY, that's one thing. Last time I looked at the protocols involved, that was the I/O model. But there's no doubt that some protocols get built over packet streams, a different abstraction. Reaching back to the early Ethernet days, some folk may recall that the original XNS protocol stack included a "Sequenced Packet Protocol" SPP, in contrast to the byte stream model of TCP. - Dave -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
