CURRENT_MEETING_REPORT_ Reported by Susan Hares/Merit Minutes of the Routing Policy System Working Group (RPS) Agenda o Logistics o Status on IRR (Daniel Karrenberg) o AS path expressions extensions to RIPE-181 (Cengiz Alaettinoglu) o Database distribution (Daniel Karrenberg) o RPSL policy term (Cengiz Alaettinoglu) o Dictionary object (Craig Labovitz) o RPSL syntax alternatives (Cengiz Alaettinoglu) o Review of work plan until next meeting All agenda topics were covered on Tuesday, 18 July, except database distribution which was covered on Wednesday, 19 July. Items are reported in order of the agenda instead of chronological order. Status Reports The following reports were given: o RA (Brian Renaud) The RA continues to cleanup the RADB from the PRDB to RADB changes. 20,000 networks have network prefixes more specific than other prefixes. The RA wants to cleanup these prefixes or mark them as withdrawn. Interactions with CA*NET and InternetMCI RRDB are in progress. o RIPE NCC (Daniel Karrenberg) Registry based in Europe. Most AS or upstream providers have put in AS maintainer objects. Consistency checking with other registries is part of the on-going work. Reports were not given for CA*NET and InternetMCI. Randy Bush asked why register in two databases? Brian Renaud stated that it is not needed at this time and MCI agrees. AS Path Expressions (See Cengiz's slides. All slides are available at ftp://ftp.isi.edu/rps/stockholm-ietf.) The focus of Cengiz's information is to allow an extended AS path syntax that allows extended AS path syntax and current syntax to work with existing tools and new tools. If both the old and the new syntax are used, the policy should be roughly the same. It was noted that the policy expressed might not be exactly the same. Tony Bates suggested using AS macros in AS path expressions and the group agreed to this. Tony and Andrew Partan noted that the AS macro expansion should expand to either the old syntax for AS path expansions or the new syntax. This new syntax has been added to RA tools. Other tool builders (e.g., Tony Bates of MCI and Marten Terpstra of Bay Networks) indicated their interest in these extended as path expressions. Any further expansion should be made with RPSL. Database Distribution Daniel Karrenberg presented the paper drafted by David Kessens and Cengiz Alaettinoglu. The problem with database distribution is that the distribution has n**2 problem with the number of registries that need to be updated. The granularity of current updates is 1/24 hours as files are FTPed from site to site. User interfaces may not be the same across registries. Daniel Karrenberg proposed: o Updating to authoritative registry The receiving registry would forward the update request to the authoritative registry for the object. If the authoritative registry is unknown, it would be bounced. o Updates will be split apart from a full database dump Each update will contain (for adds/deletes) time stamp, sequence number, and registry name. Each update will be checksummed with MD5. o Full database fetched (FTP) Each full database has time stamp and sequence number of the last update processed. The full database pull will be checksummed (MD5?) to ensure the data is correct. o Journal files will contain updates Each journal file will contain updates that can be applied after the last full database pull. Each journal updates will contain time stamp, sequence number and registry. o Asynchronous updates should be possible to send to users This service will be subscribed to. Subscribers will receive (via e-mail or other transport) the updates listed above as they occur. The group discussed how to get consistency working across databases. The focus was using maintainer objects which specified priority of registries. Tony Bates and Marten Terpstra noted that multiple maintainer objects can exist for routing objects. Since the frequence of new maintainer objects is low, time stamps and human interaction can be used in the 4 databases to resolve differences. The group recommended that the history of the databases needed to be two full copies backward plus associated journal files. Bill Manning and Paul Vixie discussed using the DNS as a Routing Registry. Weakness in DNS was: bootstrapping problem (Michael Patton), indexing and problems with the tools for consistency checking. Strengths of DNS are its hierarchical nature and the methods by which it delegates to others. Those who wish to experiment may. If databases grow, within two years we may need a hierarchical solution. Bill will find an RA draftee to experiment with a DNS based routing registry idea. Holing punches in CIDR blocks has been addressed by RIPE-127. The current code does support this. Daniel volunteered to publish this fact to the list. RPSL Policy Term (See Cengiz Alaettinoglu's slides for full a description.) RPSL policy terms have an event name, a filter, event parameters, and actions. The split of event name, filter and actions is for efficiency of processing. Five options were discussed for the interaction of different policy terms. Consensus was to go with one of the two options: specification order or a combination with the overrides rule. The combination would use the overrides rule within a policy terms and specification order between policy terms in the database. It was asked why this discussion is occurring and suggested that the group concentrate on getting the tools working with simple policy. Cengiz responded that for simple cases, policy is simple. For more complicated cases, it will aid. However, it appears that documents indicating the application will be needed for the working group. Dictionary Object (See Craig Labovitz's slides.) The dictionary provides extensibility to the language. The dictionary would contain current descriptions plus reference implementation. The documents would contain syntax description plus a place to point to for reference implementations. New changes to the RPSL would include shared libraries or perl code that could be loaded. Andrew Partan and others wondered that the Dictionary seemed two steps ahead of the current technology. He was more interested in seeing more tools with current things. RPSL Syntax Alternatives Should the RIPE-181 syntax be extended or replaced by a new syntax? In the extended RIPE-181 syntax, interas-in and interas-out syntax is replaced by an extended as-in and as-out syntax. If a completely new syntax is used, programs would just read in the new syntax elegantly. If the RIPE syntax is extended, tools using it will have to be compatible with the old syntax using hacks. Tony Bates thought it would be good to have both new syntax for the as-in and as-out plus get rid of the interas-in. At the time it was created it was thought that the separation of local and global policy was important. Review of Work Plan Until Next Meeting Items to do for the next meeting: o Daniel Karrenberg will write a registry object definition paper. o Daniel will write a document outlining his proposed approach on coordinating databases. o Cengiz and David Kessens will write a document on the consistency model that gets rid of the source attribute and uses the authoritative registry idea based on the maintainer object. There was consensus to go with this approach rather than the multiple sources approach. o Daniel will announce the non-hole punching option in CIDR block to the list. (The status field now has a PI/PA function.) o Daniel and Marten Terpstra will examine all RIPE documents and publish what changes needed to be done to them. The working group will then begin to re-write the RIPE documents as RPS documents. This change is subject to RIPE DB Working Group and the Area Director's approval o RPSL Language Cengiz, Craig Labovitz and Marten Terpstra and other Routing Registry people will continue work on this and report back at the next IETF.