CURRENT_MEETING_REPORT_ Reported by Paul Francis/Bellcore and Dino Farinacci/cisco Systems Minutes of the Inter-Domain Multicast Routing Working Group (IDMR) The IDMR Working Group met during the morning sessions on Tuesday and Wednesday. Summary o The two PIM documents (PIM = Protocol Independent Multicast, formerly ESL), dense and sparse modes, were presented and discussed. Though some details about the phase shift between sparse and dense mode need working out, the general consensus of the group is that the multiple scaling modes approach is desirable. Implementation and development of PIM will continue. o No work was done on CBT, but a status report was given describing CBT's state of implementation (almost done). There is still interest in CBT as valuable work, either as a potential alternative to PIM (if PIM proves overly difficult), or as an experimental protocol. o The group decided to propose a new name and charter to better reflect that the focus is no longer strictly inter-domain, but rather scaling versus quality in general. Paul Francis will generate the proposal. First Session The meeting started with a brief presentation by Deborah Estrin of sparse mode PIM (PIM-S). The main purpose of this presentation was to set the context for a subsequent presentation of dense-mode PIM (PIM-D) by Dino Faranaci. Dino's presentation was followed by a more detailed presentation of PIM-S by Deborah. Deborah's presentation focused primarily on the differences between the current and previous specifications. Both specifications were well received. Dino stated that he will start an implementation of PIM-D in the next few weeks. Deborah will continue implementation and simulation of PIM-S. There were two general concerns with this work. One was that there is little need for PIM-D in the absence of PIM-S, given that PIM-D does not do much beyond what current multicast (DVMRP and MOSPF) already do. The other concern was the phase shifting mechanism between PIM-S and PIM-D. This concern was not based on scepticism that it could work so much as simply the lack of specification and experience with it. Work and implementation on this aspect of PIM will continue. Dino recorded the following specific comments made during the presentations: o Dense-mode ESL comments - Issue with asymmetric routing even in dense-mode. - Interest in not requiring unicast and multicast boundary congruence. - Brought up issues on what policy decision for phase conversion. - Worried about lost packets - how long will we have black holes. o Sparse-mode ESL comments - Steve Deering commented that it scales well for the number of members per group, but it is not known if it can scale based on the number of groups. Concerning sources, he feels that clearing is a concern but unicast aggregation helps. Going from shared tree to per source trees reduces linearly. - Van Jacobson feels that source entries can be group independent. - Scott Brim is concerned about masks in reachability messages. - Deborah Estrin is concerned about switching from shared trees to per source trees. Dave Clark suggested using an interface specific case. - Eric Nordmark, before the second meeting, brought up the issue of a looping problem in sparse-mode (packets appear on the Ethernet twice, then start looping). RP / \ / \ ^ v A B ^ v | | ------------- ^ v | | C D / \ SRC Rec These two presentations were followed by a general discussion of scaling issues in multicast. There are many types of multicast applications, ranging from groups with a single or small number of sources and a high data volume (video broadcast or conferencing) to groups with potentially many senders and receivers, but with very occasional traffic, (such as a news group). Because of this wide range of applications, a range of techniques for dealing with scaling, and a means of dynamically moving from one technique to another, is required. It was generally felt that the current work goes a long way towards improving scaling according to the number of sources, but that scaling according to the number of groups is a major problem. Van suggested the use of a tree shared by multiple groups as one approach to scaling according to the number of groups. Second Session The agenda for this session was to cover two specific ideas for improved scaling by source. One was Deborah's idea of the use of source masks in PIM-S. The other was Van's idea of having border routers proxy for sources outside of a routing domain so that scaling in the domain is according to the number of border routers instead of the number of true sources. As Deborah was unable to attend, the presentation of source masks was given by Dino. It was generally felt that source masks increased complexity and introduced a scaling problem of their own, without significantly improving the scaling problem it tries to address. However, since Deborah was not there to defend it, the issue remains open. This was followed by a presentation of the proxy idea by Van. Since this presentation was not prepared in advance, and since all the details had not been worked out, no final consensus on the value of the idea could be reached. Van agreed to be responsible for seeing that a specification of the idea is written up. Attendees Susie Armstrong susie@mentat.com Jim Barnes barnes@xylogics.com Tony Bates tony@ripe.net Stephen Batsell batsell@itd.nrl.navy.mil Nutan Behki nebhki@newbridge.com Rebecca Bostwick bostwick@es.net Robert Braden braden@isi.edu Scott Brim Scott_Brim@cornell.edu Ronald Broersma ron@nosc.mil Glen Cairns cairns@mprgate.mpr.ca Ken Carlberg Carlberg@cseic.saic.com Stephen Casner casner@isi.edu Isidro Castineyra isidro@bbn.com John Chang jrc@uswest.com Ping Chen ping@ping2.aux.apple.com J. Noel Chiappa jnc@lcs.mit.edu David Clark ddc@lcs.mit.edu Michael Collins collins@es.net James Davin davin@thumper.bellcore.com Stephen Deering deering@parc.xerox.com Julio Escobar jescobar@bbn.com Deborah Estrin estrin@usc.edu Dino Farinacci dino@cisco.com William Fenner fenner@cmf.nrl.navy.mil Dennis Ferguson dennis@ans.net Robert Fink rlfink@lbl.gov Eric Fleischman ericf@atc.boeing.com Paul Francis Francis@thumper.bellcore.com Atanu Ghosh atanu@cs.ucl.ac.uk Ramesh Govindan rxg@thumper.bellcore.com Regina Hain rrosales@bbn.com Susan Hares skh@merit.edu Dimitry Haskin dhaskin@wellfleet.com Shai Herzog herzog@catarina.usc.edu Christian Huitema Christian.Huitema@sophia.inria.fr Phil Irey pirey@relay.nswc.navy.mil Van Jacobson van@ee.lbl.gov Merike Kaeo mkaeo@cisco.com Charley Kline cvk@uiuc.edu Peter Lothberg roll@stupi.se Tracy Mallory tracym@3com.com David Marlow dmarlow@relay.nswc.navy.mil Jun Matsukata jm@eng.isas.ac.jp Donald Merritt don@arl.army.mil Greg Minshall minshall@wc.novell.com Randy Miyazaki randy@lantron.com Doug Montgomery dougm@osi.ncsl.nist.gov Robert Moose rmoose@gateway.mitre.org Sath Nelakonda sath@lachman.com Erik Nordmark nordmark@eng.sun.com Vijayaragavan Pandian vjp@proteon.com Andrew Partan asp@uunet.uu.net Radia Perlman perlman@novell.com Eric Peterson elpeterson@eng.xyplex.com Ram Ramanathan ramanath@bbn.com Yakov Rekhter yakov@watson.ibm.com Allyn Romanow allyn.romanow@eng.sun.com Greg Ruth gruth@gte.com Henning Schulzrinne hgs@research.att.com Isil Sebuktekin isil@nevin.bellcore.com Ming Sheu msheu@vnet.ibm.com Keith Sklower sklower@cs.berkeley.edu Andrew Smith asmith@synoptics.com Martha Steenstrup msteenst@bbn.com John Stewart jstewart@cnri.reston.va.us Bernhard Stockman boss@ebone.net Larry Tepper ltepper@compatible.com Fumio Teraoka tera@csl.sony.co.jp Akihiro Tominaga tomy@sfc.wide.ad.jp John Veizades veizades@ftp.com Alice Wang alice.wang@eng.sun.com Chris Wheeler cwheeler@cac.washington.edu John Wroclawski jtw@lcs.mit.edu Mary Jo Zukoski maryjo@gateway.mitre.org