CURRENT_MEETING_REPORT_ Reported by Jessica Yu/Merit Minutes of the CIDR Deployment Working Group (CIDRD) Agenda o CIDR deployment status o Aggregation do's and don'ts o Request for more generality in NSFNET NACRs o CIDR block allocation guidelines o Router VLSM functionality test reports o NIC check/renumbering Administrivia This group was formerly called ``BGP Deployment and Application'' (BGPDEPL). The name has officially been changed to ``CIDR Deployment'' (CIDRD) to better reflect its current mission. The mailing list for the group has not changed. Overview The group agreed unanimously and emphaticly that CIDR needs to be deployed now. The last major piece of core BGP-4 interoperability is to be deployed on the Saturday after the meeting when ANS and the CIX will start speaking BGP-4 to each other. There are already about 100 networks being routed by aggregates alone; there is proof now that CIDR is working and that it is reasonably safe to switch to it. The next group meetings for discussing CIDR deployment will probably be the NATCOM (formerly ``Regional-Techs'') meeting in Ann Arbor 2-3 June, and the RIPE meeting. These meetings are too far in the future to defer CIDR deployment issues until then. Progress on CIDR needs to be made now; any coordination necessary to make that happen needs to happen in this meeting or on mailing lists. There was an extended discussion about whether this group should pick a date and then stop routing where providers had not aggregated. A straw poll showed 18 hands supporting such a unified cutoff date somewhere around 1 June 1994. There were some strong cautions (including from Scott Bradner, Operations Area co-Director) that IETF working groups do not have the formal structure to allow this kind of a trade decision. This might constitute legal restraint of trade. It would be better for NSPs to make such decisions independently (especially with independently selected dates), but not formally as this group. Andrew Partan offered the following perspective: as a service provider, he will monitor his routing tables and will keep the list posted; he will not set an arbitrary date now. But if the situation grows more critical, he will do what is necessary to keep his network viable. Suggestions: service providers should enable proxy for stub sites; look at ASs that are not creating aggregates, and do proxy there. CIDR Deployment Status Jessica Yu presented the following points: o Many ASs have deployed BGP-4 and are CIDR-capable now o All major providers are either ``CIDRizing'' or defaulting o About 30 ASs are advertizing about 100 CIDR routes o About 200 more specific routes have been withdrawn o The community desperately needs to further reduce routing table size o New ways to do operations in a CIDR world need to be found Tony Bates reported that twenty-seven aggregates are not announcing any more-specific routes. See the files ftp.ripe.net:cidr/stats/Specific* for daily updates. Eric-Jan Bos presented tables from The Netherlands showing: with ``classic'' measures: 19758 routes (24 As, 4143 Bs, 15591 Cs); looking to CIDR measures: four ``B'' blocks + 109 ``C'' blocks (1x/9, 4x/15, 20x/16 ... 34x/23). Only thirty-five ASNs are announcing aggregates (of 338 known ASNs)! Merit posts lists of aggregate nestings and withdrawals (also noting policy conflicts) in: merit.edu:pub/nsfnet/cidr/nestings.announced CIDR progress, as seen on AS690, is also tracked in cidr_savings in the same directory. Aggregation Do's and Don'ts Andrew Partan presented the following points: o Proxies need to know the local topology. o Sometimes when people announce a large route and withdraw the more specific routes, they forget that you still need to route some of those more-specifics (e.g., customers who have moved, multi-homed ASs). o Check outside of your routers to see what is really happening in the outside world. Meta-Aggregate NACRs Peter Lothberg requested a change in the way that AS690 handles aggregate configurations. He would like to be able to configure one aggregate, and then be allowed to announce any classless network within that aggregate and have exactly what he announces be propagated into and through AS690. This would allow him to change his aggregation strategy on the fly without registering each aggregate that he intends to use in the future with the NSFNET PRDB. Dennis Ferguson said he would look into what this would require. Merit will also review the routing stability implications of this change. CIDR Block Assignments It would be very helpful to have a document that NSPs can hand out to their customers to explain the CIDR conservation situation and rational CIDR block allocation policies. (This has come up before, but it is becoming more urgent.) Vince Fuller presented a list of some pieces that already exist: o IANA to regional NICs (i.e., RFC 1466) o Regional NIC to providers (i.e., RIPE-104) o Provider to customers (i.e., RIPE templates) o Customer block usage (subnetting document? Harvard Eidnes document extract?) How many documents exist? What is the structure of the documents? Customer block assignment issues: How big should blocks be? (The customer doesn't know.) Vince presented the following questions/issues when this question was asked: o How many host IP addresses? o How many network segments? o Hosts per segment (range)? o What type of routers? o WAN links (i.e., multiple sites in customer network)? - If yes, #1-3 for each site. - Un-numbered serial links? o Goal of >10% block utilization (Postel to SESQUInet) o RFC 1597---how much of the net is public? Charlie Kline presented his tool for managing CIDR allocations. It is available from: ftp.cic.net:/pub/src/tree.tar.Z ftp.cic.net:/pub/src/tree.hqx (Macintosh executable) Peter Ford also pointed out that we need to document CIDR and allocation guidelines for two categories of large future usage: SLIP and PPP protocols. David Conrad, Tony Bates, Marten Terpstra and Andrew Partan will work on these documents. Peter Ford will co-ordinate the editing of this document. Goals are 1 May for a draft, and a final copy after the next RIPE meeting. WIDE VLSM Report Akira Kato made a presentation of his work with Hiroshi Kawazoe on a project to determine if VLSM is ready or not. His slides follow the minutes. A test was held on 24-25 November with the following routers: IBM 611/140, mpnp.1.1.1.2; Proteon CNX-500, V15.0a[Z1]; cisco 3000, 9.1.8; 3Com NetBuilder II, XW6.2.0.10; and Sony NWG-5000WSN, NEWS-OS 6.0. Product performance is not identified by router in the final report, by agreement with the participants. Most routers passed most of the tests they performed, except one which failed to handle supernets. Conclusion: VLSM is almost ready. Action Items o NSPs leave Seattle, go home and withdraw some routes! o Use bgpd@merit.edu to make CIDR progress known. o NSPs will press their customers and peers to implement CIDR. o NSPs will start to proxy-aggregate stub-ASs. o RIPE and Merit will set up CIDR tool repositories for public use (ftp.ripe.net:/cidr/fdata,docs,stats,toolsg). o Group to produce guidelines documents (details above). Attendees Nashwa Abdel-Baki nashwa@frcu.eun.eg Vadim Antonov avg@sprint.net Susie Armstrong susie@mentat.com William Barns barns@gateway.mitre.org Tony Bates tony@ripe.net Jordan Becker becker@ans.net Steven Blair sblair@dell.com Erik-Jan Bos erik-jan.bos@surfnet.nl Scott Bradner sob@harvard.edu Ronald Broersma ron@nosc.mil Brad Burdick bburdick@radio.com Jeffrey Burgan jeff@nsipo.nasa.gov Joesph Burrescia burrescia@es.net Randy Bush randy@psg.com Henry Clark henryc@oar.net Michael Collins collins@es.net David Conrad davidc@iij.ad.jp Steve Corbato corbato@nwnet.net Sean Doran smd@use.net Tom Easterday tom@cic.net Havard Eidnes havard.eidnes@runit.sintef.no Nasser El-Aawar nna@ans.net Erik Fair fair@apple.com Steve Feldman feldman@mfsdatanet.com William Fenner fenner@cmf.nrl.navy.mil Dennis Ferguson dennis@ans.net Robert Fink rlfink@lbl.gov H. Tom Fitzpatrick fitz@ddn.af.mil Vince Fuller vaf@barrnet.net Dimitry Haskin dhaskin@wellfleet.com Eugene Hastings hastings@psc.edu Kenneth Hays hays@scri.fsu.edu Denise Heagerty denise@dxcoms.cern.ch Steven Hubert hubert@cac.washington.edu Jinho Hur jhhur@cosmos.kaist.ac.kr Geoff Huston g.huston@aarnet.edu.au David Jacobson dnjake@vnet.ibm.com Dale Johnson dsj@merit.edu Matthew Jonson jonson@ddn.af.mil Merike Kaeo mkaeo@cisco.com Akira Kato kato@wide.ad.jp Hiroshi Kawazoe kawazoe@trl.ibm.co.jp Sean Kennedy liam@nic.near.net Edwin King eek@atc.boeing.com Charley Kline cvk@uiuc.edu Mark Knopper mak@aads.net John Krawczyk jkrawczy@wellfleet.com Tony Li tli@cisco.com Lars-Johan Liman liman@sunet.se Kim Long klong@nysernet.org Peter Lothberg roll@stupi.se Jamshid Mahdavi mahdavi@psc.edu Bill Manning bmanning@rice.edu Matt Mathis mathis@psc.edu Jun Matsukata jm@eng.isas.ac.jp Keith Mitchell keith@pipex.net Pushpendra Mohta pushp@cerf.net Gilles-Andre Morin gamorin@shl.com Ngoc-Lan Nguyen lnguyen@icp.net Peder Chr. Noergaard pcn@tbit.dk Donald Pace pace@cntfl.com. Krishnan Parameshwaran krishnap@microsoft.com Kurt Parent kurt@nwnet.net Andrew Partan asp@uunet.uu.net Michael Patton map@bbn.com David Piscitello dave@corecom.com Rex Pugh pugh@hprnd.rose.hp.com Yakov Rekhter yakov@watson.ibm.com Robert Reschly reschly@brl.mil Tony Richards richards@sprintlink.net Francois Robitaille francois.robitaille@crim.ca Duncan Rogerson d.rogerson@nosc.ja.net Michal Rozenthal michal@fibronics.co.il John Scudder jgs@merit.edu Tim Seaver tas@concert.net Henry Sinnreich hsinnreich@mcimail.com Bernhard Stockman boss@ebone.net Tim Streater t.c.streater@dante.org.uk Marten Terpstra marten@ripe.net Paul Traina pst@cisco.com Willem van der Scheun scheun@sara.nl Ruediger Volk rv@informatik.uni-dortmund.de Chris Wheeler cwheeler@nwnet.net Linda Winkler lwinkler@anl.gov Cathy Wittbrodt cjw@barrnet.net Philip Wood cpw@lanl.gov Jessica Yu jyy@merit.edu Paul Zawada Zawada@ncsa.uiuc.edu