Editor's Note: These minutes have not been edited. RPS WG Meeting Minutes (Wed 10:15am - 11:15am) Reported by Ramesh Govindan Cengiz Alaettinoglu started the meeting by describing changes to the RPSL specification. The first change involves adding a new exception specification to the as-in statement. This doesn't change the expressivity of the language, but makes it convenient to express exceptions, and makes it less error-prone. Exceptions may be arbitrarily nested. A second change is in support for tunnels. For this, we need to modify the ifaddr statement to add an identifier for a tunnel. To the tunnel object, we added a way to specify the tunnel encapsulation mechanism, and a way to specify the tunnel protocol. Changes to the community attribute include allowing a list of communities to be specified with the "contains" method. Also changed the next-hop attribute to allow assignment to "self". Cengiz asked whether the document could be forwarded for IESG consideration. No one present dissented. Dave Meyer talked described the Internet-Draft "Application of RPSL on the Internet". This document explains how RPSL might be used to specify policy objects: the document describes examples of common objects, some canonical policy examples, and the use of tools. The document also discusses a preferred naming convention for AS sets or AS macros and other high-level objects. Dave presented several examples described in the I-D. There was confusion about the use of <> in one of the examples: Randy Bush and Alan Barrett explained the difference between the use of <^AS3582$> and AS3582 in the policy filter. Cengiz suggested including both examples in the draft. Randy suggested explaining in greater detail the semantics of regular expressions used in policy filters, specifically in the provider-provider policy description. Some discussion in the RtConfig description about cisco's AS path length limitation: Tony Bates indicated that the 255 character limit had been fixed. Dave also suggested adding to the draft some text about how to manage an aut-num object. Cengiz said that the recommendations specified in the applications draft would be implemented in aoe. One question was whether roe worked on the RIPE database: RIPE doesn't have an updated whois server, but will soon have accouding to Chris Fletcher. In the meantime, Cengiz suggested the use of the RA whois server, which mirrors RIPE data and can update RIPE registry. David Kessens talked about the transition from RIPE to RPSL. A modification to the RIPE database for RPSL is well underway. The transition plan involves allowing RIPE-181 syntax for updates as long as possible, as well as have tools recognize both syntaxes. The transition plan has four stages, each stage taking approximately two months. Allow RPSL use now, but initially have RIPE-181 as default. Stage 1 tests server software: in addition, tutorials at Nanog and RIPE to encourage use. Stage 2 integrates RPSL capability in tools. In Stage 3, run two databases in parallel, with the old format server on the old port and the RPSL-capable server on a new port. In addition, at this time, old data will be converted to RPSL format. In Stage 4, RIPE and RA will switch their servers to RPSL. Tony Bates asked whether 2 months per stage is realistic, particularly because some providers have their own tools and internal extensions. Tony suggested having hard timelines, otherwise in his experience from transitioning from Ripe-81 to Ripe-181, things never move on. Randy pointed out that other providers will also have to update their tools before transition happens. Tony also said that, to give providers incentives to switch, the document should be forwarded onto the standards track. RPSL implementation in RAToolSet As the last talk, Cengiz presented a tools update. Specifically, some of the tools have been made RPSL-capable. These changes are now in alpha-test. Changes include: more specific route operators, aspath prepending, community attibutes (lacks operator methods). Also, as a transition mechanism, if a policy object has rpsl-in/out, tools will use that as opposed to as-in/out (which may be in RIPE format). Before tools can be used, need database support for syntax checking.