Minutes of 6bone BOF (draft version 0.2 7Apr03)
18 March 2003
San Francisco, IETF-56
Jordi Palet <firstname.lastname@example.org> chaired the meeting; The minutes were formed by notes taken by Tim Chown, Jordi Palet and Mat Ford.
Attendance was ~150.
Web site: <http://www.6bone.net>
Status and discussion of the proposal for the “transfer of 6bone address management responsibilities to RIRs” - Bob Fink, 30 mins
Presentation of the 6bone phase-out plan - Bob Hinden, 10 mins
Discussion of the 6bone phase-out plan - 20 mins
Status and discussion of the proposal for the “transfer of 6bone address management responsibilities to RIRs” - Bob Fink
Bob Fink briefly overviewed the 6bone's history. The 6bone has been running for seven years as an operating testbed, and the ipng/ipv6 project has started 10 years ago. The original purpose of the 6bone was to test standards and implementations, and the original address allocation came from RFC 2471 by Fink, Hinden and Postel. Over the years the focus moved towards testing operational procedures and helping to evolve the IPv6 production prefix allocation procedures, with 6bone experience as a bootstrap.
Current general use is for early experience, for free, for ISPs and organizations to see if IPv6 is for them. However, routing performance and stability is not always the best (e.g., see the Savola 6bone-mess I-D), so we need a more robust core that can support testbeds. Some pTLA allocations have recently been reclaimed.
In early 2002 discussions were started with the RIRs driven by two issues: Clarifying the role the 6bone address registry has with respect to the RIRs IPv6 address registry and gaining access to the ip6.arpa reverse registry.
Not at issue, contrary to common perception, is that it was being done because Bob Fink was retiring as Bob always intended to stay involved to assist in any appropriate way, or that the IETF ngtrans wg was closed thus a new home was needed. The IETF still feels responsible for the 6bone
During the course of early discussions, the RIRs’ management made it clear that they could not speak to the issue of how long the 6bone allocation authority would last. Rather, it was an issue that the IETF and/or the IANA would have to deal with. To this end, a discussion was opened within the IETF on 6bone phaseout planning.
Also there was the issue of would oversee operations? Even though the 6bone came under the IETF ngtrans wg, it really had almost nothing to do with its operational policies. The 6bone community itself controls its policies and everyone expects that it will continue to do so.
Many comments came from the 6bone community, but most relevant ones focus on having to pay for testbed addressing when they haven’t had to pay in the past. Note that many 6bone participants (at all levels) do so to get experience and in the process convince their organizations that there is something worth paying for, i.e., the price is an issue, no matter how small it is.
There was concern about having to go through more complexity. It isn’t clear if this is a real issue as we don’t know what a pTLA-level request process might be with the RIRs. This may be a holdover of dislike of necessary procedures for scarce IPv4 address space.
Another concern is what is pay for service when the 6bone is a volunteer effort… RIR services aren’t needed. There is unwillingness to pay for service and then be expected to hand our free address services to downstream users
Many comments have come from the RIR community as well. Bob said that he would paraphrase the single largest concern as “…why should the 6bone community get cheaper services than the dues paying members…?”
Also, the RIRs are supposed to recover costs for providing their services. Giving away any service would seem to go against this. A corollary to the above is, if the RIRs are just covering costs for a special service to the 6bone, what are the RIRs doing for their regular customers. The feeling seemed to be, why should RIR members care about the 6bone? Let 6bone do their thing, and the RIRs theirs.
Bob made it clear that the above are his readings of the concerns and comments, not any RIR presentation of them.
It isn’t clear this proposal should proceed, given the opinions expressed on both sides, a soon to be in place 6bone phaseout plan, a decline in the request rate for 6bone prefixes, and a steady increase in allocated production prefixes.
Also, there is now the ability for the RIRs to temporarily allocate IPv6 addresses for Internet experiments.
As for e.f.f.3.ip6.arpa support, Bob has proposed that the 6bone operate the servers for this themselves, which would mean that the 6bone community would sustain the cost of entering and maintaining the pTLA data in the e.f.f.3.ip6.arpa server., and that when phaseout is complete, the RIRs simply pull the eff3.ip6.arpa delegation and they have reclaimed it.
The RIRs have agreed that in light of the foregoing that there is no need to continue planning for a 6bone RIR integration, and that the 6bone would continue to manage its own allocations throughout the life of the phaseout plan.
The RIRs will delegate e.f.f.3.ip6.arpa to name servers that the 6bone community provides.
Michel Py agreed with Bob's interpretation of the opinions expressed on the 6bone and RIR lists with regard to this. He also thinks that transition to RIR would prolong life of 6bone and this is not desirable, therefore supports the conclusions of Bob's presentation.
Geoff Huston noted that a special considerations letter was sent from RIRs to IAB regarding providing ip6.arpa support for 3ffe, and that it is not up to RIR to delegate e.f.f.3.ip6.arpa to the 6bone - it should be done by the IANA.
Randy Bush said that the formal way to make this happen is an I-D with IANA considerations to request the IAB to ask IANA to delegate this way. Randy said he will help Bob do this.
Presentation of the 6bone phase-out plan - Bob Hinden
Bob Hinden then presented the 6bone phase-out plan I-D, under the subtitle "be careful what you start". RFC2471 says 6bone addresses would be temporary addresses that would be reclaimed in the future (with implied renumbering for sites using 6bone addresses). The RIRs have been allocating production SubTLAs since 1999.
Bob's personal view is that the 6bone should continue until people stop using it. The currently observed decline in requests for 6bone address space, and growth of RIR production allocations, indicates that 6bone is starting to go away naturally.
In 2002 more production allocations were made than 6bone ones. The v6ops WG replaced ngtrans, which used to oversee the 6bone, but the 6bone is not in the v6ops WG Charter.
The current plan outlined in the draft is to allocate 6bone addresses until July 1, 2004, and for these to remain valid until July 1st 2006, after which no 6bone prefixes should be carried on the Internet. The plan obsoletes RFC2471. It is up to IPv6 address holders to gain new address space, of whatever prefix length is appropriate (some pTLA holders may only require a site prefix, for example). In addition, IANA must not reallocate 3ffe:/16 for at least two years to avoid confusion with new allocations.
There has also been a suggestion for Jan 1, 2005 and July 1, 2005 as the allocation cutoff and phaseout/shutdown dates for the 6bone. Also that the IANA reuse text that comments on the list said change "MUST make provisions to set it aside..." to a SHOULD.
Discussion of the 6bone phase-out plan
Michel Py noted that there is is a distinction for 6bone between general experiments and original "new stuff" experiments. We should shutdown soon, but there is no good or bad date.
Tim Chown noted that routing stability comes by behavior not a specific prefix. Some 6bone is native, some production nets are tunneled. He felt that the original timeline is reasonable as there is no barrier to getting production address space, but the more interesting discussion is about getting stability on 6bone.
Michel Py said there is no point having a 6 month cutoff, so Jan/July is not right, we need at least 1 year from cutoff to kill date.
Bob Hinden asked if we perhaps should consider 1 July 2004 and 1 July 2005 as the stop allocation and phaseout dates respectively?
Mat Ford said that the IANA re-use should be a MUST - people are going to install filters on the shutdown date and any ambiguity will cause problems.
Bob Fink noted that this has been discussed - people are opposed to saying 'filter now' - want to allow flexibility for IANA to re-use this address space as they see fit.
Mat Ford felt that we need to be clear on the filtering date and its implications.
Bob Hinden then asked if we should use July 1 2003, and July 1 2005?
Tim Chown asked why not use 06/06/06 for the kill date as it had a nice symmetry?
David Kessens stated that he had no preference on the window size (between stop allocation and phaseout), but we must be clear on the implications of the plan for users. Also that too many options will make it harder to get consensus.
Pekka Savola said that his experience with renumbering is that it can take a year, so we need to give time to move. We should stop allocating as soon as possible, but give more delay on phaseout/shutdown.
Bob Fink asked if it is better to shutdown in the summer rather than (Xmas) holiday time? Comments seemed to say that summertime was better.
Marc Blanchet stated that the final date is most important.
Christian Huitema felt that Jan 1, 2004 is OK (we wont get an I-D through by July 2003).
Jordi Palet noted that you can you can get an experimental prefix from RIRs, you just need an upstream provider to forward the request on your behalf.
Tony Hain said that he gets a lot of "How do I get a pTLA?" questions. I tell people they must go and get production address space.
Anne Lord (APNIC) noted that there is a charge for an experimental prefix from APNIC which is our lowest
possible fee ($625 US per annum), allocated on a one-year basis, and that you can apply directly.
Axel Pawlik (RIPE) said that to get experimental space you must become a member. He noted that some 90% of 6bone users in the RIPE region are already members.
Ray Plzak (ARIN) said that there is no ARIN policy yet on experimental addresses.
The meeting voted for cutoff and kill/shutdown dates and chose:
Allocation cutoff: 1/1/04
Bob Fink will reissue the phaseout draft with the changes from this meeting.
The meeting adjourned.