Le 4 août 2010 à 05:21, Brian E Carpenter a écrit : > On 2010-08-04 00:49, Rémi Després wrote: > ... >> In my understanding, a reason why it is usually set to 0 is that a stateful >> operation, which is complex, is so far mandatory (because a pseudo-random >> number has to be assigned to each flow, not a stateless hash). >> If a 5-tuple hash is permitted in the future, stateless operation will >> become possible, and more and more hosts can be expected to set the field. > > No, that can't be the reason,
Not "the" reason, just "a" reason ;-). > because RFC 3697 sets no strong rules - > it doesn't even forbid just counting 1,2,3... as successive labels. The quote is: "To avoid accidental Flow Label value reuse, the source node SHOULD select new Flow Label values in a well-defined sequence (e.g., sequential or pseudo-random) and use an initial value that avoids reuse of recently used Flow Label values each time the system restarts". Right, well-defined sequences aren't all pseudo random, and there is only a SHOULD. But it does implicitly advise to avoid 5-tuple hashes (they don't produce well-defined sequences). Right? Tony Hain told me that, to his recollection, the reason why hashes were not clearly permitted was privacy protection. This seems to me an overkill in the name of privacy. Explicitly leaving the choice between stateless and stateful would make the rule less strong in favor of statefulness. It would make it more practicable because of the inherent simplicity of statelessness. RD > At the time, nobody wanted to set any rules. > > Brian > -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
