Le 2013-02-08 à 11:59, Havard Eidnes <h...@uninett.no> a écrit : >>> However, there is nothing which enforces RFC4291-conforming IIDs for >>> (for instance) statically configured IPv6-addresses. >>> So in what way do >>> well defined u/g values for RFC4291-conforming IIDs help you? >> >> 1. >> Users that manually configure their IPv6 addresses should be >> knowledgeable about what they do. >> >> If one configures an address has u=g=1, unless it is for an >> experiment, it is a human mistake (it conflicts with the IPv6 >> addressing architecture of RFC 4291). Depending on the context, >> this mistake will have no consequence or will eventually have >> to be corrected. > > What's the operational failure mode? Does it fail in a way which > would be obvious to the ones who has made this mistake, or will > simply some seemingly-random pairs of hosts fail to communicate? > I think it would be an engineering mistake to willfully create > hard-to-detect failure scenarios via a modification to already > fielded standards.
FULL AGREEMENT that we don't want to modify any "already fielded standard". If you believe such a change is needed for 4rd, PLEASE EXPLAIN. If one configures an address that doesn't follow configuration constraints, it may fail in various ways, but under its own responsibility. For instance, if it has the wrong subnet prefix, it won't be reachable from across the Internet, like what may happen if it has a non-RFC4291-complying IID that happens to be in the 4rd reserved range in a site where 4rd is activated. > >> Note that, even if one does such a mistake in a 4rd site, the >> likelihood of using the 4rd IID prefix 0x0300 remains low. > > That's a nice handwave... This is a fact, worth noting IMHO, especially since the particular example of a non RFC4291-compliant address Ole gave didn't conflict with 4rd. Your lack of interest for this point doesn't justify, IMHO, a so depreciating comment. >> 2. >> On my home network, I know nobody is doing manual configuration. > > That doesn't really prove anything. It doesn't prove anything, OK, and wasn't BTW claimed to prove anything in particular. Yet, the risk of manual misconfigurations that may conflict with the RFC 4291, which is indeed a risk, is mitigated by the fact that manual address configuration is AFAIK more an exception than a general practice. Regards, RD > > Regards, > > - Håvard -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
