On Fri, 2007-12-07 at 10:32 +1300, Brian E Carpenter wrote: > On 2007-12-06 13:28, Per Heldal wrote: > > and ULA > > does IMHO not qualify as such. > > IMHO ULA does qualify, in fact must qualify, since ULAs > have technical impact (see my previous note and Tony Hain's > comment on it). >
I still consider it more of an operational issue than a technical one. It's naive to think that every standard can be made such that there are no limits. Users, operators and vendors have to be able to deal with it. You wrote: >Another way to say it is that there is a default expectation >that ULAs will be filtered and that PI prefixes will be >routed. That's a good enough rationale for having them in >separate parts of the address space - a ULA being routed >or a PI prefix being filtered are both things that should >raise an operational flag, and that will be easier if they're >distinguishable at a glance. This doesn't make it an IETF issue. No policy-making body worthy of operating in this field would be short of people who understand this issue. As a comment to your statement Tony wrote: >Actually there is, because the public network operator community >-thinks- they know what the important/easy issues are for enterprise >network operators. Never mind that they have no real clue about how >difficult it is to carve up portions of an otherwise aggregated space >and keep that correct at all instances over an extended period of time. >Clear separation is a great simplifier, and should be good enough, but >repetition of this conversation by people that want nothing more than >to tell others how to run their network shows it is not. This assumes that policy-making organisations are dictated by network operators. That is simply false. I don't intend to start a discussion of which policy-making body is more democratic. It is however relevant to mention that the organisations in the ASO represent a combination of individuals, SMBs, large enterprises, academic communities and network operators. Governments and national regulators have also started to pay attention and take part in the processes in recent years. > To go further, PI should qualify too, since the allocation of > PI poses a serious technical threat to the scaling of the > routing system. As does an explosive growth in PA blocks being assigned, new TE schemes causing massive fragmentation, and a number of other things. If IETF needs detailed control of the routing table's size, you have to address those policy-issues too. For the record; I don't oppose any particular definition of ULA. I just don't want them as IETF standards. It's dangerous to generalise, but standards are often hard-coded or even implemented in silicon, while policies at best form a base for default configurations. Regardless of where the decision is made it'll eventually end up as an entry in IANA's list of allocated address-ranges. Any of a number of suggested methods for an id/locator-split has the potential of solving the scalability issue -- at least for edge networks. That'll largely take care of PI, and even the dreaded ULA-C(G) may then be routed globally with no impact on the DFZ should anyone feel like doing so. //per -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
