Thanks for the great discussion Brian. I think we’re all in sync now? Dino > On Aug 30, 2018, at 8:46 AM, Brian Trammell (IETF) <ietf@xxxxxxxxxxx> wrote: > > > >> On 30 Aug 2018, at 16:55, Dino Farinacci <farinacci@xxxxxxxxx> wrote: >> >>> On Aug 30, 2018, at 2:57 AM, Brian Trammell (IETF) <ietf@xxxxxxxxxxx> wrote: >>> >>> hi Dino, >>> >>> Almost. How about: >>> >>> >>> OLD: >>> >>> When the UDP and LISP headers require integrity protection, the >>> methods of using UDP checksums in [RFC8085] can be considered. >>> >>> NEW: >>> >>> Implementors are encouraged to consider UDP checksum usage guidelines in section 3.4 of [RFC8085]. Specifically, when the UDP, LISP, and outer IPv6 headers require protection against corruption, the use of non-zero UDP checksums is RECOMMENDED. >> >> Well if we recommend it and when describing the UDP header in the packet format section we don’t that woudl be a contracdiction. > > I think my point here is that the packet format section probably shouldn't do that. :) Yes, I understand the disconnect between the reality of the situation and the > >> And note the IPv6 outer header cannot be protected with a UDP checksum. The link-layer CRC will do that. > > Eh, this makes assumptions about the underlying link layer's corruption characteristics that may not hold. But yeah, for most packets in most realistic situations this is the case, and I guess we've learned to live with the underlying phy error rate * 1e-10 in any case. > >> NEWNEW: >> >> Implementors are encouraged to consider UDP checksum usage guidelines in section 3.4 of [RFC8085] when >> it is desirable to protect UDP and LISP headers against corruption. >> >> What do you think? > > This seems like a fine compromise to me. > > Thanks, cheers, > > Brian > >> >> Dino >> >> _______________________________________________ >> Tsv-art mailing list >> Tsv-art@xxxxxxxx >> https://www.ietf.org/mailman/listinfo/tsv-art >
