In your previous mail you wrote: > => short prefix lifetimes are not the common case: prefix lifetimes > are at least in months and are announced in days. And don't forget > the 2 hour rule. I agree that in the common case there are no 'short' lifetimes, but I wouldn't call it "months"
=> in fact this is more in years: I changed our prefixes twice in 5 years. (default is 7 days for preferred and 30 days for valid). Especially I doubt that RAs are "announced in days" by default. => 7 days and 30 days are in days, aren't they? Still there are circumstances when short lifetimes are necessary, e.g. during a renumbering event or when abusing RAs for router redundancy. => I can't parse the second case. This all is not my point. My point is, that an interface doesn't try to reconfigure itself when it loses its address. => a global address is not useful without a default router. > => no but it can react to the lost of all default routers (default > lifetime is 600 seconds, maximum is 1800 seconds). I agree, that might also be possible. But a quick check showed that neither Linux, BSD nor Windows hosts send RSs' upon Router Lifetime timeout either. => my point is that they could. But this advise is also lacking in RFC2462. If I searched it correctly, the RFC does not talk about Router Lifetime at all. => yes, everything about Router Lifetime is in RFC 2461. And still, when the prefix' valid lifetime is shorter than the Router Lifetime (Maximum is 18.2 hours? 16bit?), there are times when the => 1800 seconds = 30 minutes. interface is lacking a global address. => IMHO it is a bad idea to add text in RFCs about cases that should never happen in the real world, in particular in standard track RFCs which are already far too large/complex. Regards [EMAIL PROTECTED] -------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
