Dear all, how about this (thanks to Henk): Definition: A physical network infrastructure (layer 2) operated by a single entity with the purpose to facilitate the exchange of Internet traffic between Internet service providers. The number of Internet Service providers connected should at least be three and there must be a clear and open policy for others to join. Addresses needed for other purposes (e.g. additional services provided to the members) should be assigned by upstream ISPs. Assignment size: - a /48 in most cases - a /64 if it is known that there will only be one subnet (basically following the IAB recommendation) Additional warning: The RIRs should warn the IXP that it is strongly discouraged to announce the addresses and that they will likely not to be globally routable. If this is agreeable the RIPE NCC will implement this policy and co-ordinate it with the other Regional Internet Registries. Mirjam Kuehne RIPE NCC
wording quibbles:
Addresses needed for other purposes (e.g. additional services provided to the members) should be assigned by upstream ISPs.
s/assigned by upstream ISPs/acquired through other appropriate means/ could be that the ix operator could qualify for normally allocated space all by themselves.
- a /48 in most cases - a /64 if it is known that there will only be one subnet (basically following the IAB recommendation)
i suspect the 'most cases' is the /64, a single exchange mesh, and that the multiple interconnected meshes, which whould need the /48, is a rare case.
The RIRs should warn the IXP that it is strongly discouraged to announce the addresses and that they will likely not to be globally routable.
the ix does not annouce mesh address. isps on the mesh may. so i would say the rirs should warn the ixp that isps usually do not announce mesh space to their peers, so that the ix space allocated under this policy is likely not to be visible globally. randy
Randy Bush writes:
[Things looking good to me]
The RIRs should warn the IXP that it is strongly discouraged to announce the addresses and that they will likely not to be globally routable.
the ix does not annouce mesh address. isps on the mesh may. so i would say
The address space is assigned to the IXP, so the IXP has authority to disallow any announcement. Thus IMHO it's sufficient to address the IXP. Robert
Randy Bush writes:
- a /48 in most cases - a /64 if it is known that there will only be one subnet (basically following the IAB recommendation)
i suspect the 'most cases' is the /64, a single exchange mesh, and that the multiple interconnected meshes, which whould need the /48, is a rare case.
Well, I'd think it's more the case that a /64 is "single now, and never plans to need more" and /48 is "likely to need some more in future", thus shifting "most cases" towards /48. Robert
- a /48 in most cases - a /64 if it is known that there will only be one subnet (basically following the IAB recommendation) i suspect the 'most cases' is the /64, a single exchange mesh, and that the multiple interconnected meshes, which whould need the /48, is a rare case. Well, I'd think it's more the case that a /64 is "single now, and never plans to need more" and /48 is "likely to need some more in future", thus shifting "most cases" towards /48.
i think of an ix as more like an isp than an end user enterprise. we are careful to allocate to isps based on reality, not hallucinations from the marketing department. randy
Hiya Randy et al, On Fri, 29 Jun 2001, Randy Bush wrote: ->i think of an ix as more like an isp than an end user enterprise. we are ->careful to allocate to isps based on reality, not hallucinations from the ->marketing department. Actually using 13 bits out of 128 for IPv6 (a /35 subTLA with a /48 minimum customer assignment) for a (global) ISP´s reservation, architecture and hierarchy, when customers get 80 bits for a dozen or less machines would seem to be based on hallucination and certainly not reality. It is a good thing the IETF say they are revisiting this - I just hope the outcome *is* based on reality. Could Randy (and others) please indicate why he appears to insist on the smallest possible assignment, at the risk of later needing either a multiple announcement or renumbering ? I find it hard to think that IPv6 addresses are in short supply when we have decided on a /48 mimimum assignment to end users. Cheers Dave
Hi, On Thu, Jun 28, 2001 at 08:22:44PM +0200, Mirjam Kuehne wrote:
how about this (thanks to Henk): [..]
I like it (for the number of reasons already mentioned in the lists and in the WG meeting). Gert Doering -- NetMaster -- SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
Mirjam Kuehne wrote:
Definition:
A physical network infrastructure (layer 2) operated by a single entity with the purpose to facilitate the exchange of Internet traffic between Internet service providers. The number of Internet Service providers connected should at least be three and there must be a clear and open policy for others to join.
I think this is as good a definition of an IXP as we will get.
Addresses needed for other purposes (e.g. additional services provided to the members) should be assigned by upstream ISPs.
I still believe this principle compromises the operational neutrality aspirations of any well-managed IXP (see my previous posting), and it is very important as Randy proposes that there is an alternative means for an IXP to get IPv6 space for these purposes in its own right (e.g. as an LIR as per IPv4).
Assignment size:
- a /48 in most cases - a /64 if it is known that there will only be one subnet (basically following the IAB recommendation)
Sounds plenty.
Additional warning:
The RIRs should warn the IXP that it is strongly discouraged to announce the addresses and that they will likely not to be globally routable.
So long as this as a consequence of how the ISPs announce the space, and not due to any property of (other than I guess prefix length), or policy restrictions adhering to, the space itself, I don't have any particular issue with this, and we'll look at encouraging our customers to adopt good practice on peering mesh announcements. Keith http://www.xchangepoint.net/contact/keith/
participants (6)
-
Dave Pratt -
Gert Doering -
Keith Mitchell -
Mirjam Kuehne -
Randy Bush -
Robert Kiessling