CURRENT_MEETING_REPORT_ Reported by Mark Laubach/Hewlett-Packard Minutes of the IP Over Asynchronous Transfer Mode Working Group (ATM) Monday The first session opened with a formal announcement by Robert Hinden that he has stepped down as the ATM Working Group chair and that Mark Laubach has assumed the responsibility. The agenda was presented and approved. A review of recent ATM Forum activities was presented by Steve Willis. He reported that the User Network Interface (UNI) Specification Version 3.0 document is expected to be ratified in August. An overview of the European ATM pilot project was presented by Juha Heinanen. The topic of ``routing IP over the switched virtual cloud'' was presented by Joel Halpern, and he volunteered to write a proposal. Consensus is that the ATM Working Group will host the proposal, but actual work will be moved to another group that deals with routing over large public networks. A general discussion was held to collect comments on Randall Atkinson's Internet-Draft, ``Default IP MTU for use over ATM AAL5 Services.'' The author was not in attendance. The last order of business was discussion of Mark Laubach's ``Classical IP and ARP over ATM'' Internet-Draft (henceforth called ``Classical''). Discussion and consensus building continued over the next two meetings. Tuesday The second session opened with discussion of a timetable of ATM activities for the rest of 1993. Both the Bellcore and Naval Research Laboratory (NRL) reference signaling codes will become available in late August or early September. Both implementations will be ATM Forum UNI 3.0 compliant, with the exception of point-to-multipoint. An IP over UNI 3.0 document is expected to be completed and have implementation experience by the November IETF meeting. The rest of the session was spent on discussion of Classical. During the discussion, the Internet Area Director, Stev Knowles, made it perfectly clear that Classical was not complete until ARP and IP multicast were fully addressed. (The position that area directors may delay an Internet-Draft from being submitted into the standards process was supported by the IAB in an open meeting later that evening.) Document review continued with a renewed sense of focus. LLC/SNAP was adopted by consensus as the default (the minimum required that implementors must support) IP encapsulation method. The IP MTU default size of 9180 octets was also adopted by consensus. Wednesday The last session opened with congratulations to Juha Heinanen for the publication of RFC 1483, ``Multiprotocol Encapsulation over ATM Adaptation Layer 5.'' Work then continued on Classical with the discussion of PVC support. A section on PVC support was generated for the document by an ad hoc team, and the text was approved by the group. An edited version of the text will be included in the document. Further discussion on Classical took place following a presentation by Mark Laubach on a solution for ARP using an APR server. The group eventually reached consensus on the solution. Mark also presented solutions for the treatment of IP broadcast and IP multicast in ATM. These were also approved. Having reached consensus on all issues, discussion on Classical was closed. Mark will produce a rewrite within the next two weeks. Juha Heinanen led a discussion on his ``NBMA Address Resolution Protocol (NBMA ARP)'' Internet-Draft. Much discussion was generated on this topic, but unfortunately not enough time was available to conclude all issues. Juha will meet with others in the working group to resolve outstanding issues. The following are detailed summaries of the various discussions including consensus decisions by the working group. -------------------------------------------------------------------------- ATM FORUM Update, Steve Willis, Mike Goguen, Andrew Malis, Joel Halpern, Drew Perkins, Mark Laubach, et al. o Signaling was closed at next meeting of the Forum in June. Touch up of point to multipoint addressing will be done in July. The ATM-FORUM will take a vote in August to adopt Uni 3.0 as an Implementation Reference. o Physical, agenda for settling issues. Time schedule: - 7/93, pick a bit rate for UTP3 (25 vs 51Mbps) - 9/93, pick a line encoding o Private NNI working group is starting in July. VC routing to be worked on in the ATM-FORUM. Mike Goguen (and probably Joel Halpern) will keep IETF experts involved where possible. Joel will likely create an information sharing activity between the ATM-FORUM working group and the IP routing over large public data networks activity (see below for more information on IP routing issues). o LAN Emulation, starts next meeting. Keeping Novell, bridging, et al. working. May be host services emulation. We've heard a rumor that they may be looking at encapsulation issues. Also, the FORUM Working Group has not decided their plans in detail. o ATM FORUM intends to support the output of the RFCs from this working group unchanged. o The Issue of getting ATM FORUM documents was raised. The Interop ATM-FORUM address was distributed and we've told folks that the Uni 3.0 spec should be available for $25.00 sometime in/after August. Mark Laubach also committed to seeing if we can find an electronic mechanism for distributing on the Internet. o Mark Laubach will contact Glenn Estes (Bellcore) regarding strengthing the information flow between the technical committee of the ATM-FORUM and this working group. Our working group time frame indicates that the November IETF meeting will likely discuss IP over UNI 3.0 standardization and any implementation experience we've gained at that point. An invite will be put to the ATM-FORUM to see if any signaling technical people can come to the working group meetings at the November IETF. A challenge will be put to the ATM-FORUM to allow IETF working group attendees to go to ATM-FORUM meetings, we believe that the FORUM's rules will not allow this. The best we can probably hope for is to have IETF working group attendees who are ATM-FORUM members to support information exchange. ------------------------------------------------------------------------ EUROPEAN ATM PILOT, Juha Heinanen Juha presented a quick look at an ATM project in Europe: o ATM is quite a big thing in Europe, bigger than SMDS or Frame relay o At least 34 Mbps pilot network o 17 network operators have sign a memorandum of understanding o Access speeds not defined in this pilot, operators can use whatever speed to get to the customers o Only the NNI is specified for PVC (virtual path) Conforming with some European standards, small subnet of CCITT spec, Ok with ATM-FORUM UNI 2.0 specification. o No more than three hops (operators) between end points. o Goal is for operators to gain experience and test the standards The real issue is that the operators want to get into the ATM bandwagon o EC competition rules would make this network illegal for the long term operation o Nordic area is aiming at 155 Mbps trunks ------------------------------------------------------------------------ IP ROUTING over the Switched Virtual Cloud, Joel Halpern Joel led a discussion of IP routing over large switched public data networks. He is preparing a proposal. As this is an IP routing issue and not an IP-over-ATM issue, further work on this will not take place in this working group. Whatever activities will take place a future IETF meetings will stay closely linked to the ATM Working Group. Points from Joel's talk: o It is not ARPs problem to figure out who you really should talk to. This applies not just to ATM, but to frame relay, and x.25 o BGP next hop is very handy o Picking up where directed ARP and short-cut routing left off. o This should be a generally applicable solution that darn well ought to work on ATM. o Can point-to-multipoint change the solution space? Joel thinks not as things should be point-to-point based. o Clearly you don't want the routing data to be non-aggregated o This came up with IDRP, can build stub-routing entities o Without a way to route over the cloud. o Juha: some sort of route query protocol where a terminal attached to an ATM network and set up a route request query to a server and get a response back. o This is not completely new work. Some ability to query and store information. Can invent a new protocol. o We want to have it before the large ATM cloud comes into existence. o We don't want to wait until IPng. o This effort will tie to the routing protocols. o Joel will create a proposal and will distribute on the mailing list A nub of a design. He will try to get a proposal out to the e-mail list in the very near future. ------------------------------------------------------------------------ MTU Draft Comments These are merely comments collected at the working group meeting as we had a large collection of people there. These comments do not represent any formal opinion of the group. o Drew Perkins: ATM FORUM terminology has changed AAL5 PDU size is 64K-1. Minimum size should be deleted from the document IP has a minimum reassembly size is 576 bytes. This is not the real minimum size. Bob: our documents should have rough description of how to reduce the MTU size..... o Juha: too much implicit stuff going on in document. We clearly need to use exactly the same mechanism is specified in the FORUM. ------------------------------------------------------------------------ Working Group Schedule The following time schedule for our working group activities was discussed. 1993 1994 ITEM M J J A S O N D | J F M A M ---------------- ---------------------------------------- Encapsulation x Classic Document x..x The "next" document x.....? ATM Forum UNI 3.0 x..x NRL Signaling x..x Release Bellcore Signaling x..x Release Framework WORKING GROUP TODOs: --------------------- 1. IP encapsulation negotiation via UNI 3.0 signaling 2. MTU size negotiation via UNI 3.0 signaling (Ran's document) 3. TCP/UDP Port mapping directly on to VCs, architecture impact 4. Routing over the Switched Cloud 5. Multicasting 6. NBMA The hopes are that with the release of the NRL and Bellcore signaling stacks, the working group should be able to review implementation experience at the next meeting in Houston. The "next" document, i.e. IP over UNI 3.0, should be reviewable by the next meeting. No one volunteered to write this yet..... ------------------------------------------------------------------------ Classical IP and ARP over ATM draft discussion items and decisions. All decisions were reached with clear consensus by the working group: o PVC support in the classical document was an issue. A section was generated by an ad hoc team during the Wednesday lunch break. The working group approved the text. An edited version of the text will be included in the classical draft. o Last part of paragraph on ANSI ITU-TSS....stricken (from introduction) o Working group approved by the default (required) implementation of 9180 bytes MTU size. Text regarding minimum size was stricken. o Working group approved that LLC/SNAP be the default (required) encapsulation for IP packets: i.e., all implementations MUST be able to support LLC/SNAP as one of the encapsulation choices. o Working group approved by the ARP server architecture model as proposed by Mark Laubach. We had some lengthy discussion on the issue of providing primary and backup servers and the working group clearly decided that a single ARP server will be required per logical IP subnet and that this would be sufficient for the near future (year) until ATM multicast or highly reliable ARP servers are implemented. The proposed model will roll to either future implementation without changes to the host. The issue was raised of soliciting the ATM-FORUM for the allocation of a well-known ATM address for ARP. o Working group concluded that current ATM standards and technology do not provide any broadcast mechanisms and as such the classical draft will not specify an IP broadcast to ATM broadcast mechanism. Hosts may transmit packets that select the IP broadcast (all ones) or subnet broadcast (all ones in host portion). Hosts, upon receiving an IP broadcast or IP subnet broadcast for their logical IP subnetwork, must process the packet as if addressed to them directly. o Working group concluded that current ATM standards and technology do not provide any multicast mechanisms and as such, the classical draft will not specify an IP multicast to ATM multicast mapping. The working group agreed that current IP multicast implementations (i.e., MBONE and IP tunneling) will continue to operate over ATM based logical IP subnets if operated in the WAN configuration. Furthermore, the working group would like to have a statement added to the IP multicast section stating something to the effect that, when ATM multicast is available, roll-over from to the new architectures will be straightforward. Mark will prepare the new version of the draft and distribute it within two weeks. As we are trying to fast track this document, technical review and final consensus on the draft will be collected via e-mail. NBMA draft review. Juha Heinanen Unfortunately, discussion of the classical draft and related issues took up most of the time of the working group. We managed to close and give 20 minutes on the last day to Juha to lead the discussion of his NBMA draft. Clearly this was not enough time as much discussion was generated. I was able to record the following comments during the discussion: o Just use source routing (Brian Carpenter), o Dennis Ferguson has issues about this should really be a routing issue and not an ARP issue and that we really should have a routing protocol that does all (in the IP layer). o Joel Halpern stated that he is thinking about this in his routing protocol proposal. Are all NBMA servers IP routers? Joel feels that we need to be able to follow the NBMA model and resolve via ARP. o Dennis would really like this issue to be solved with an IP level protocol. SMDS has a different ARP mechanism than ATM, but this NBMA issue is the same. Dennis would like to have a media independent solution. Dennis wants a cleaner separation. o Mark Laubach would like a clean description of the changes to the routing decision process / architecture on a host (when it makes decisions and what gets relaxed). o Juha is getting together with Joel to work on the issues. Attendees George Abe abe@infonet.com Roland Acra acra@cisco.com Masuma Ahmed mxa@sabre.bellcore.com Kannan Alagappan kannan@DSMAIL.ENET.DEC.COM Arun Arunkumar nak@3com.com Cynthia Bagwell cbagwell@gateway.mitre.org Nutan Behki Nutan_Behki@qmail.newbridge.com Lou Berger lberger@bbn.com Vincent Berkhout berkhout@cs.utwente.nl Carsten Bormann cabo@cs.tu-berlin.de Michael Brescia Caralyn Brown cbrown@wellfleet.com Tracy Brown tacox@mail.bellcore.com Theodore Brunner tob@thumper.bellcore.com Steve Buchko stevebu@newbridge.com John Burnett jlb@adaptive.com Ramon Caceres ramon@mitl.research.panasonic.com Brian Carpenter brian@dxcern.cern.ch Les Clyne l.clyne@jnt.ac.uk Jonathan Davar jdavar@synoptics.comm Kurt Dobbins dobbins@ctron.com Jeffrey Dunn dunn@neptune.nrl.navy.mil Tom Easterday tom@cic.net Ed Ellesson ellesson@vnet.ibm.com Robert Enger enger@reston.ans.net Julio Escobar jescobar@bbn.com Mark Fedor fedor@psi.com Dennis Ferguson dennis@ans.net James Forster forster@cisco.com Osten Franberg euaokf@eua.ericsson.se David Fresquez fresquez@vnet.ibm.com Dan Frommer dan@jeremy.enet.dec.com Shoji Fukutomi fuku@furukawa.co.jp Eugene Geer ewg@cc.bellcore.com David Ginsburg ginsb@us-es.sel.de Mike Goguen goguen@synoptics.com Ramesh Govindan rxg@thumper.bellcore.com Marcel Graf graf%dhdibm1.bitnet@vm.gmd.de Ron Greve rgreve@cs.utwente.nl Joel Halpern jmh@network.com Patrick Hanel hanel@yoyodyne.trs.ntc.nokia.com Ken Hayward crm57d@bnr.ca Geert Heijenk heijenk@cs.utwente.nl Juha Heinanen juha.heinanen@datanet.tele.fi John Hopkins J_Hopkins@icrf.icnet.uk Jeff Hughes jeff@col.hp.com Sascha Ignjatovic sascha@veda.co.at Phil Irey pirey@relay.nswc.navy.mil Ronald Jacoby rj@sgi.com David Johnson dbj@cs.cmu.edu John Johnston john@berlioz.nsc.com Peter Kaufmann kaufmann@dfn.dbp.de Lothar Klein lothar.klein@gmd.de Mark Laubach laubach@hpl.hp.com Mark Lewis Mark.S.Lewis@telebit.com Carl Madison carl@startek.com Andrew Malis malis_a@timeplex.com Allison Mankin mankin@cmf.nrl.navy.mil Jun Matsukata jm@eng.isas.ac.jp Keith McCloghrie kzm@hls.com 3 Donald Merritt don@arl.army.mil Topi Miettinen tm86214@cs.tut.fi William Miskovetz misko@cisco.com Daniel Myers dan@nsd.3com.com David O'Leary doleary@cisco.com Masataka Ohta mohta@cc.titech.ac.jp Zbigniew Opalka zopalka@agile.com Charles Perkins perk@watson.ibm.com Drew Perkins ddp@fore.com Roy Perry rperry@advtech.uswest.com Philip Prindeville philipp@res.enst.fr J. Mark Pullen mpullen@cs.gmu.edu James Reeves jreeves@synoptics.com Tony Richards richards@icm1.icp.net Benny Rodrig brodrig@rnd-gate.rad.co.il Hal Sandick sandick@vnet.ibm.com Tim Seaver tas@concert.net Henk Sennema sennema@sara.nl W. David Sincoskie sincos@thumper.bellcore.com Timon Sloane timon@timon.com Kenneth Smith kensmith@bnr.ca Michael St. Johns stjohns@darpa.mil Antoine Trannoy trannoy@crs4.it Catherine Treca Catherine.Treca@dione.urec.fr Hisao Uose uose@tnlab.ntt.jp Dono van-Mierop dono_van_mierop@3mail.3com.com Werner Vogels werner@inesc.pt Scott Wasson sgwasson@eng.xyplex.com James Watt james@newbridge.com Jost Weinmiller jost@prz.tu-berlin.d400.de Marcel Wiget wiget@switch.ch Kirk Williams kirk@sbctri.sbc.com Steven Willis steve@wellfleet.com Rachel Willmer rachelw@spider.co.uk Sam Wilson sam.wilson@ed.ac.uk Paul Zawada Zawada@ncsa.uiuc.edu