On Jun 4, 2013, at 11:11 PM, Lorenzo Colitti <lore...@google.com<mailto:lore...@google.com>> wrote:
On Wed, Jun 5, 2013 at 12:02 PM, Ted Lemon <ted.le...@nominum.com<mailto:ted.le...@nominum.com>> wrote: So then your argument should be "RIRs should not plan to assign /48s to subscribers because ISPs are assigning /56s to subscribers anyway"? No, it shouldn't. My argument is that the belief that no bits are available for use in semantic prefix-based routing is not sustainable. Wait, but the email I just replied to was talking about user allocations. I guess the question is: if every user gets a /48, are there still bits available for semantic prefixes or not? If so, then we don't have to have this conversation. If not, then it seems to me that the situation is that ISPs can choose to either assign users /48s or use semantic prefixes, but not both. No, that's not at all a central point of this debate. Someone said that RIRs won't give ISPs more prefix than they need to give each of their customers a /48. It has been pointed out that RIRs do not in fact have a hard-and-fast policy to this effect, so in fact even an ISP that gives out a /48 to every customer may be able to use semantic prefixes, depending on the specific policies of their "local" RIR. In addition, I pointed out that in fact RIR allocation policies based on the _assumption_ that each end-user site would get a /48, in combination with the general tendency of ISPs to _actually_ allocate only a /56, mean that there are eight bits to play with even if the RIR policy is quite strict. The ISP can legitimately argue that they are giving those bits to the customer, because they are assigning multiple prefixes to the customer. So there simply isn't a problem getting bits if the ISP decides this is a solution they want to implement. If that's the case, you can certainly then say "there's no point in giving users /48, it's too much" - that's a perfectly valid opinion to hold. However, we must take into account that today, RIR policy is based on allowing ISPs to assigning /48s to users. Which of the two is it? There is no such dichotomy. The question of how wide a mask we ought to give end-users isn't really even open for debate. RIRs are not, at the moment, imposing policy on ISPs that restricts them to only one /56 per customer. And it's generally agreed (perhaps not by you) that a /56 is going to be plenty for a typical end-user home network. So the point isn't that a /48 is a waste of space. It's that a /48 is assumed, and because it is assumed, there are definitely bits available for semantic prefix assignment.
-------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------