CURRENT_MEETING_REPORT_ Reported by Guy Almes/Advanced Network and Services Minutes of the CIDR Deployment Working Group (CIDRD) Tony Li's Report on IP Address Space Usage and Lifetime Tony summarised the IPv4 address space and how the InterNIC allocates large CIDR blocks. Further, he noted how the InterNIC maintains the statistics on allocated and assigned address space used in this report. Given this background, Tony issued the disclaimer that he could not estimate possible non-linear effects due to future developments. He then proceeded to discuss his estimates of future Address Space Usage and the Lifetime of the IPv4 Address Space. At the San Jose IETF (December 1994), Tony had estimated a lifetime of the year 2008 +/- three years. By the Danvers IETF (April 1995), the growth slope appeared to drop. The new projected lifetime is the year 2018 +/- eight years. The statistical extrapolation technique is Tony's visible eye-balling. Tony then presented the implications for the development and deployment of IPv6. From his IPv4 projections, the worst-case date by which a deployed IPv6 is needed is the year 2010. From his assumptions about IPv6 development and deployment, the latest that an intense development-and-deployment effort could begin would be seven years total (the critical path is two years for host development plus five years for host deployment). Thus, this intense IPv6 effort would need to begin in the year 2003. Tony's slides are included at the end of this report. Frank Solensky's Report on IP Address Space Usage and Lifetime Frank presented his slides on projections similar to Tony's. However, his techniques for statistical extrapolation was more sophisticated. One slide showed the growth in 128/2 (i.e., Class B) allocations. The estimate here projected that, under current allocation policies, the 128/2 space would not run out. A second slide showed the growth in 192/3 (i.e., Class C) allocations. There were two distinct allocations here (labeled raw and smoothed). The raw estimate showed 192/3 not running out, but the smoothed estimate showed 192/3 running out by the end of the decade. Frank's slides are included at the end of this report. Erik-Jan Bos's Report on Routing Table Growth Erik-Jan reported on the growth in the number of prefixes present in routers. The number of total nets continues to grow, and at a higher rate than we would like. Similarly the number of AS numbers continues to grow. The number of CIDR routes continues its strong upward growth, as has the number of AS numbers doing CIDR. 56% of the AS numbers now advertise at least one CIDR prefix. Erik-Jan has been maintaining a database, with entries for each hour since January 1994, of the number of BGP entries in Amsterdam1.dante.net. This plot shows an uncomfortably high slope since Danvers (April 1995). The proliferation of IP providers, for example, leads to the creation of holes in SURFnet's CIDR blocks. Tony Bates stressed that we need to continue to put pressure on the `top 20' providers who need to improve their use of CIDR. Erik-Jan's slides are included at the end of this report. Daniel Karrenberg's Report on Routing Table Growth Daniel reported on the number of prefixes for each 8-bit prefix. His table includes the number of host addresses possible (given its CIDR breakup), the number of actual current routes, and the number of hosts per route. Daniel has done this once per month for the last two months. The value of this report will grow as he continues to produce these reports each month in the future. Bill Manning's Report on Class A Space Utilisation and Class A CIDRisation Bill reported on the state of allocation of the traditional Class A space and lower (i.e., not reserved) 0/1 space. The IANA is negotiating to recover some of the pre-CIDR allocated 0/1. About 3% of the total IPv4 space has been recovered in the last few months. They are continuing to review nets not visible in the global routing system, and hope to recover an additional two to four additional 0/1 prefixes. Bill's slides are included at the end of this report. Classless in-addr.arpa Delegation -- Geert Jan de Groot Geert Jan addressed the problem of maintaining the in-addr.arpa entries of the DNS when different CIDR blocks of a single class-full network number are assigned to different organizations. After briefly discussing two non-solutions, he presented an admittedly ugly solution that works with current software. Specifically, use CNAME RRs to alias/move authority to another zone. This has actually been on the net for a while and seems to work well. Havard Eidnes and Geert Jan have written an Internet-Draft on this technique. This technique is also being discussed within the DNSIND Working Group. Geert Jan's slides are included at the end of this report. Address Allocation for Private Internets -- Yakov Rekhter Yakov noted that private internets are proliferating, and that the use of global IP addresses for these private internets depletes the address pool. To avoid this, we need to provide an alternative addressing plan. To further this, Yakov has drafted a successor to RFC 1597 that: o Discusses those cases when non-global addresses can be used, o Specifies three specific blocks that can be used for this, o Notes implications for the global Internet routing system, and o Apologises for the likely need for private internets to renumber. During discussion, it was noted that DNS records sometimes expose an otherwise private IP address. In some cases, this will require two overlapping sets of DNS definitions for a company's IP domain. As a postscript, Yakov noted the need for administrators to renumber and for tool-builders to help automate renumbering. The key argument is that, for a site's IP addresses to be useful, it does not suffice for them to be globally unique. Many sites are likely to have to renumber lest, for example, their old globally unique prefix be taken back by IANA (to recover a valuable and underused Class A, for example) or their old globally unique prefix be dropped from the routing tables of providers (to recover routing table memory space, for example). During discussion, the point was made that renumbering should be applicable across the entire range, including providers and very prestigious large sites. What is needed is routable addresses -- not merely unique addresses. In a discussion of operational considerations, it was brought up that if a firewall tries to support proxy Web service, and if the Web browser tries to use Web authentication, then this fails. It was judged that this is a HTTP protocol problem. It was also brought up that Tony Bates has already written an Internet-Draft that calls for a reserved set of AS numbers that can be used for private Internets. In discussion, Eliot Lear warned that RFC 1597bis needs to be done with care, to avoid negatively impacting users. Yakov's slides are included at the end of this report. Standards-Track Actions o Best Current Practices Scott Bradner reported that Dave Crocker believes that the current standards process supports what Best Current Practices were intended to achieve. Dave thus objects to the notion of Best Current Practices. About a dozen CIDRD members had read the proposal for Best Current Practices, and none found it objectionable. The Operational Requirements Area would like to do things such as RFC 1597bis without running it through the standards track. Scott advised that, if members feel that the Best Current Practices is useful and important, then they should so advise the IESG. o The Address Ownership Internet-Draft Yakov and Tony Li have written an Internet-Draft on address ownership, taking the position that a notion of address ownership has and will cause the Internet to be unroutable. It is likely that a working group last call on this Internet-Draft will come soon. o An Appeal to Return Address Space Internet-Draft Phil Nesser has written a Internet-Draft urging people to return unneeded address space, but he was not present at the meeting. Bill Manning has agreed to help with any editorial rewrite. Tony needs to ping Phil on the status of this document. o Net 39 Experience Geoff Huston's document should be edited for shorter sentences, but then sent ahead as an Informational or Best Current Practice RFC. Barry Greene volunteered to help edit it. o Help in IP Space Renumbering There has been a request for someone to write a document on what is involved in renumbering an IP network. This is important in order for people to sanely agree to ever renumber their networks. Geert Jan volunteered to write a draft. Sean Doran's experience is that most customers are happy to renumber when the importance of doing so it pointed out to them. There was discussion of the proliferation of IP addresses due to the common practice of multiple Web server domain names, supported by single server machines, requiring multiple IP addresses due to mis-features of the HTTP protocol. Scott Bradner reported that the HTTP Working Group is working on this problem.