Dear All, Sorry for cross posting this message to more than one IPv6 related mailing list, but I think IPv6 community should prepare with some position to the transit/exchange only providers. I have seen the excellent presentation from Mirjam K�hne from RIPE-39 about IPv6 exchange proviers, but then I have not seen any discussion. My goal with this problem statement to increase the debate and hopefully reach some consensus about the problem. The problem statement is rather long.... Sorry about that. Kind Regards, Janos Mohacsi Problem statement for transit/exchange only IPv6 address space: The current Proposed TLA and NLA Assignment Rules (RFC 2450) and IPv6 Aggregatable Global Unicast Address Format (RFC 2374) is not particularly supporting addressing long haul transit providers or exchange providers. As RFC2374 states: "While this address format is designed to support exchange-based aggregation (in addition to current provider-based aggregation) it is not dependent on exchanges for it's overall route aggregation properties. It will provide efficient route aggregation with only provider-based aggregation." What is the current proposed solution to obtain IPv6 addresses for long-haul IPv6 transit providers or IPv6 exchange providers? To better understand the case I try to explain the problem more thoroughly. Let suppose the following situation: There is a long-haul ipv6 transit provider call it X. X is providing IPv6 transit or exchange service for ISPs Y1, Y2,....Yn. But X provide only IPv6 address assignment service for few customers of Yi, let say only for 4-5 customers, for one of the following reasons: 1. Most Y1,Y2, ... Yn has already production IPv6 subTLA 2. X company profile does not contain running local IP registry. 3. other reason X operates its offices in the area of Yi. For office purpose X can obtain IPv6 address from Yi. For IPv6 transit provision X needs an IPv6 address, but probably not /35 addresses, but /44 is enough, even if the transit network is quite extensive and geographically large. Scenario 1: X request address for its transit network from Yi, since his office is in the area of Yi. They get /44 from Yi (it must be difficult according to the current NLA allocation rules). X is implementing transit service and establishes private or public peering with ISP Z1, Z2, ... Zn and of course ISP Y1, Y2, ... Yn. To help the network management X implement and run some monitoring tools on the network near to monitored equipment. X can access its network, the involved network was network Yi. They are happy with the monitoring tools and X wants to share some results with Zk and Yj. Since the strict aggregation network addresses: of X-transit_network will be reacheable via Yi peering of Zk and Yj, not via peering of X-Zk and X-Yj that would be preferred. X can advertise more specific routes Zk and Yj, but this way X pollutes with non really aggregated entries the routing table of Zk and Yj. If they implement a really strict aggregation scheme then these more specific routes are discarded. How can X solve the problem? What happens if X wants to change provider from Yi to Yj? What happens if X has more than office one near to Yi and the other Yj? Which one should be used? Or some kind of multihoming should be used? Scenario 2: X request address (a subTLA) for its transit network from Regional IP Registry. X implementing IPv6 transit service and also network monitoring as before, but he does not allocate addresses just for few ISP Y. This somehow violates the current registration policy, since X does not fulfill all the 5.1 requirements of RFC 2450 and the corresponding RIR IPv6 allocation policy documents. There is no problem with accessing the services via the preferred peering since aggregation is done in X based on subTLA. Is this the current recommended solution? Scenario 3: X request address for office purpose only from Yi, and for the network they use site local address. The public and private BGP peering is quite problematical in this case, since some routers does not allow using local type IPv6 addresses in BGP peering. Providing the monitoring services can be also difficult, since monitoring tools cannot be reached outside of X transit network. This solution is not really acceptable. Or you have some other idea?
----- Original Message ----- From: "Janos Mohacsi" <Janos.Mohacsi@dante.org.uk> Problem statement for transit/exchange only IPv6 address space: The current Proposed TLA and NLA Assignment Rules (RFC 2450) and IPv6 Aggregatable Global Unicast Address Format (RFC 2374) is not particularly supporting addressing long haul transit providers or exchange providers. ... This solution is not really acceptable. Or you have some other idea? The "toy" IPv4 Internet is a sewer. IPv8 is designed to be a swamp to cover the sewer. IPv16 is the "high-ground".... ...here are some links... Jim Fleming http://www.unir.com Mars 128n 128e http://www.unir.com/images/architech.gif http://www.unir.com/images/address.gif http://www.unir.com/images/headers.gif http://www.ntia.doc.gov/ntiahome/domainname/130dftmail/unir.txt http://msdn.microsoft.com/downloads/sdks/platform/tpipv6/start.asp http://www.ietf.org/mail-archive/ietf/Current/msg12213.html http://www.ietf.org/mail-archive/ietf/Current/msg12223.html
Hi, I haven't seen any responses to this, so let me comment... On Fri, Nov 09, 2001 at 10:09:35AM +0000, Janos Mohacsi wrote: [..]
Problem statement for transit/exchange only IPv6 address space:
The current Proposed TLA and NLA Assignment Rules (RFC 2450) and IPv6 Aggregatable Global Unicast Address Format (RFC 2374) is not particularly supporting addressing long haul transit providers or exchange providers.
I don't think there is any issue that needs to be addressed. There are three different cases: - the ISP in question is offering IPv6 address space to its customers, that is, will do LIR services. In this case he can get address space from the RIRs, no problem. - the ISP in question is not offering LIR services, but has an upstream provider who *is* offering LIR services. So he can get an /48 from them with no problem. A /48 means "a full /64 for 65000 networks", which is quite some. If /127s are used on point-to-point links (which is legal) it's definitely "enough for ever". If /64s are used, it may be necessary to come back to his upstream provider and get a second /48 - which is possible. No problem here either. - the ISP in question is a Tier-1 provider, has no upstream provider that can provide IPv6 addresses to him, and does not offer LIR services to his customers. (Out of curiosity: does such an ISP exist? I haven't found one.) If such ISPs exist in reality, they could get an /48 from the "Yi" customer that you mentioned - or they could just become a LIR and get address space for their own, which is not impossible, and the costs for becoming a LIR should be neglectible for a Tier-1 provider. No big problem. Conclusion: I do not see the need for a special-case provider who needs IPv6 addresses, has no upstream provider to get a /48 and does not want to become a LIR. Gert Doering -- NetMaster -- Total number of prefixes smaller than registry allocations: 72980 (73128) SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
participants (3)
-
Gert Doering -
Janos Mohacsi -
Jim Fleming