On Fri, Mar 28, 2014 at 10:48:55PM +0100, Robert Raszuk wrote: | Hi Hannes, | | > since most v6 hardware implementations i am familiar with, do not have | > support for parsing SRH processing, what you are suggesting here | > *is* a migration to a different dataplane. | + | > those are by no means "minor extensions" - those are "major, intrusive | > changes which often require hardware re-spins and expensive upgrade | > cycles" | | I do not think this is a question of build in support for parsing in | hardware SRH. | | I am of the opinion that this is a question of having hardware | complaint to RFC6564 recommendation of IPv6 Extension Header Format | which has been around for few years as IETF standards track document. | | Clearly if your hardware is not complaint with recommended format of | any new EH your claim is true. | | However it seems to me (no matter how great IPv4+MPLS data plane | feature richness is today) that freezing support for current and | future enhancements in IPv6 is a dead end street for both vendors and | operators.
i am certainly interested in *new, non-redundant* functionality. however not leveraging technology which has "multi protocol" in its name and as such serves v6 perfectly well, just for the sake of inventing something "new" does not seem prudent to me. | I guess it is ok for market to decide, but it does not seems to be ok | for IETF politics to make such call. agreed that customers buying kit are making a call on this, not IETF - however IETF should keep honesty in the discussion. unfortunately the current way of promoting v6-sr is an 'oversell'. "look we have this new dataplane, it comes at (almost) zero cost, and is going to be the 'holy-grail of networking' (one protocol, one dataplane which does it all)" so a bit more honesty on the one-off and perpetual cost of adding dataplane #4 on the internet would be appreciated. | > just curious - whats wrong with BGP-LU, BGP-LS as a NNI protocol ? | | Nothing wrong, but those are limited to a same trust domain. While BGP | SAFI 2/1 is a bit of much wider deployment scope. not sure i understand - why is 3107 limited to the same trust domain ? /hannes _______________________________________________ spring mailing list [email protected] https://www.ietf.org/mailman/listinfo/spring
