On Mar 3, 2008, at 12:09 PM, Geoff Mulligan wrote: > While a agree with Phil, I think that Pascal does have an interesting > suggestion in using one of reserved bits to indicated that the UDP > checksum has been elided because the protocol being used provides > protection of the same headers and data with at least the same > strength. > It would then be required by any bridge/router that is expanding the > packet to calculate the UDP checksum and insert it.
Right. I'm not trying to argue that none of these optimizations are worth it. Rather, I'm just trying to point out that on paper you see their benefits but not their costs, and so we should keep the costs in mind. The UDP checksum versus message authentication codes does boil down to very different error resilience, however. I haven't seen experimental results for 15.4, but experimental results for 802.11 from MIT indicate that many bit errors are in short bursts[1], something CRCs are better at catching than cryptographic operations are. Not all check bits are equivalent in their error detection properties. Phil [1] Allen Miu et al., "Improving Loss Resilience with Multi-Radio Diversity in Wireless Networks," MOBICOM 2005. _______________________________________________ 6lowpan mailing list [email protected] https://www.ietf.org/mailman/listinfo/6lowpan
