Brian E Carpenter <[email protected]> wrote: > On 2011-09-27 03:38, John Leslie wrote: > >> Alas, Brian, this strikes me as a rather bad idea -- introducing >> ambiguity and almost inviting gaming the system so that the same >> packet looks different to a DPI device and the actual destination. > > I don't understand that comment. There's no ambiguity in the > packet - it may contain the option, but if it does, the option > isn't ambiguous.
Isn't it? What if it points somewhere other than the actual "end of the header chain"? Doesn't that make the packet look different to middleboxes that use this option? (I'd rather not, of course, argue whether that makes the _option_ ambiguous -- IMHO it certainly makes _something_ ambiguous.) > Any node is at liberty to ignore it, including a DPI device that > wants to inspect all the headers (which a paranoid firewall > presumably would). Absolutely -- but that doesn't "cure" ambiguity. :^(` >> IMHO, any node interested in such details of the payload should be >> prepared to run at wire speed -- with a dedicated pipeline element if >> anyone were so foolish as to perform this in the "backbone". > > Why? It's a design tradeoff, and the option suggests a different > tradeoff point. IMHO, tradeoffs which introduce ambiguity are _bad_ tradeoffs. (YMMV, I suppose, but I _really_ hope it doesn't!) -- John Leslie <[email protected]> -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
