CURRENT_MEETING_REPORT_ Reported by Greg Minshall/Novell Minutes of the IP Routing for Wireless/Mobile Hosts Working Group (MOBILEIP) Thanks to Pierre Dupont for taking notes for these minutes. Greg Minshall provided opening remarks and a brief history of the MOBILEIP Working Group. Charlie Kunzinger gave a short presentation on the current Mobile IP Draft. A question and answer session followed the presentation. o Q: Why not two IP addresses for MH? A:(Charlie Kunsinger) No need for two addresses A:(Steve Deering) MH can acquire pop-up address to act as its own FA o Q:(Tony LI) Does FA decrement TTL in IP header before forwarding message to MH? Will this interfere with traceroute and MH location privacy? A: general discussion ensued on security requirements and the pors/cons of TTL being decremented by FA. Issue was left for further discussion on mailing list. o Q: How do two hosts with same subnet address communicate (one local, other mobile)? A: Proxy-ARP can be used to resolve addresses o Q: why not use source routing instead of tunneling? A: too many problems with source routing, so it was agreed in NJ to use encapsulation o Q:(Phil Karn) Can an MH be registered with more than one FA at the same time? This would allow MH to use either FA, and prevent continuous registration flip/flop between FAs when MH is on a cell boundary. A: general discussion followed, with no clear consensus on whether this would be beneficial. For further discussion on mailing list? o Q:(Yakov Rekhter) Draft document should be clear about how mobile IP breaks the IP subnet model. A: Deferred for later discussion. o Q: Why use IP for registration protocol? why not use UDP? A: Discussion on 'architectural purity' vs ease of implementation followed. Some IP implementations do not provide an IP interface, while all have a UDP interface. Deferred for further discussion. o Q: Can Yakov expand on subnet model question? A:(Yakov Rekhter) The IP over Shared Media draft addresses similar problem. The traditional model assumes that only hosts with same subnet address can talk directly to each other. Mobile IP means that some hosts with same subnet ID cannot communicate directly. Also, how do mobile hosts with different subnet IDs but on same physical subnet communicate? o Q: Request that an authorization type be included before all authorization fields in mobile IP messages. A: Agreed. o Q:(Tony Li) Question on Incarnation number in Agent Advertisement message. Some MH may not have non-volatile storage. Also, how is it used? A:(Dave Johnson) It is so that visiting MH can tell if FA has crashed, an therefore if it must re-register with FA. o Q: Why not use Internet Security Protocol? A: No decision has been reached on this yet. Adopt a wait and see attitude with respect to IP security. It is not the mobile-ip wg's job to solve IP security problems. A suggestion was made to not include any security fields in mobile IP messages. o Q: Are timer values defined? A: The units and field sizes are defined, but not the recommended values. There may be dependancies between timers that need to be considered. o Q: How and when does HA advertise reachability by proxy-arp? A:(Andrew Myles) HA should never advertise unless its a router also. A HA that is not a router uses proxy-ARP to intercept messages for MH. A discussion followed on whether the HA should always be a router. o Q: Would like to see characteristics and behavior of HA included in draft. A: Agreed. o Q:(Steve Deering) When tunneling to FA, what happens when the MH is not being served by the FA? Does packet go back to HA? A: IMHP deals with this. o Q: If address resolution mechanism is not ARP, there may be a problem using proxy-ARP. o Q: Why wait for a Home-Foreign confirm before sending notification to the prior foreign agent? A: The new FA is not authorized to serve the MH until it receives the confirm message from the HA. A message to the prior FA may not be required in this case, since the HA will direct messages to the new FA as soon as it has authorized it, therefore there is no need for the old FA to inform prior FA (the HA can inform the prior FA, after it has authorized the new FA). IMHP Draft Andew Myles gave a presentation on the IMHP draft. Topics included: o A definition of the MH, FA and HA elements. o The HA configuration (i.e., HA is not necessarily a router). o A new element, the cache agent, which keeps track of [MH, FA] bindings. o Security (rationale for weak security). o Home subnet communication (performance requirements, routing options). o Notification to the prior FA. On this final point it was mentioned that notification to the prior FA must be fast so that it does not become a black hole for packets. The protocol should allow the new FA to accept packets from the prior FA before the MH is authorized to use the new FA. The MH must inform the prior FA as soon as it moves to a new FA. A period of questions and answers followed. o Q:(Steve Deering) How are loops eliminated? A: A number of alternative mechanisms exist to break routing loops. o Q:(Steve Deering) How does routing work when a FA crashes? (Black Hole) A: A timeout will occur on cache entries, causing polling to the destination FA (the Cache Agent polls the FA every timeout secs) o Q: How does Cache Agent get bindings? A: Snooping can be used for dumb hosts. This can be turned off in the Cache agent is desired. o Q:What if MH moves from an authorized FA to an unauthorized FA? The MH will be temporarily using an unauthorized FA. A: During discussion it was pointed out that the FA may want to bill someone (HA) for the service to the MH. Therefore the new FA may not want to provide service to the MH until it is authorized to do so by the HA. o Q: The Cache Agent may send redirect packets to any host. This could compromise security/privacy (e.g., location information). A: A flag could be used to prohibit route forwarding o Q: What about ad-hoc networking? A: for further study o Q: The cache timeout/polling mechanism may generate too much network traffic. A: Polling would only occur when the route is "active". Outstanding Issues Charlie Kunzinger presented a list of outstanding issues for discussion. o Encapsulation method. Generic or Home-grown? We need at least one required method. Steve Deering argued against negotiation. Tony Li mentioned there already exists an Internet-Draft on encapsulation (Generic Routing Encapsulation). Dave Johnson stated that it had a large overhead and may not be compatible with ICMP (in terms of header size). Yakov Rekhter stated that GRE was already implemented and being deployed. Steve Deering stated that generic encapsulation can be used with a reason encoding (e.g., Mobile IP host). Greg Minshall recommended that the group continue discussion on the mailing list and pick an encapsulation method later. o Foreign Agent receives forwarded message to MH for which it has no binding. What does it do with the message? This issue was discussed at the last session. o Should address fields be expanded to include address type and length? Steve said that it may depend on how often packets are sent. Dave said the protocol is IP specific, address must fit into 64 ICMP bits and Tony recommends addresses be TLV fields to support multi protocols (e.g., Mobile appletalk). No consensus was reached. o Do we need to control the number or frequency of registration requests? A discussion followed on whether to allow MH to register in multiple cells (i.e., with more than one FA) and have HA duplicate messages to both FAs. Steve suggested that protocol should not disallow this, but recommended it be deferred to the advanced functionality issue list. This issue was left unresolved. o Is there a need for a retransmission timer on a registration request by the MH? It was suggested that the MH be allowed to retransmit a request and that the FA could respond with an in-progress message if it is awaiting a response from the HA on a previous request for the MH. o State diagrams in draft document? This will be included in the next revision. o Should the protocol allow a hierarchy of HA? Should not preclude this option in draft. o Can TOS bit in IP header be used to identify mobile hosts? Dave stated that RFC 1122 suggests this is not possible. o Why can an FA terminate service to an MH? Also, HA can deregister MH. It was suggested that there is no need to include FA to MH deregistration since it will time out eventually. o Several comments were made on the style, packet format and byte alignment in the draft. o Should ICMP or UDP be used for registration protocol? After some discussion, a poll was taken on the preferred method and UDP was selected by a majority of those responding. o Weak security: definition needs to be included in the draft. o To what degree do we break the subnet model? This is similar to the problem with large public data networks (e.g., ATM). Yakov volunteered to communicate to the IAB how Mobile IP will break the subnet model (and write an Internet-Draft?). Cache Agent Model A discussion on the pros and cons of the intermediate Cache Agent model followed, with no consensus being reached on how to proceed. Some argued it should be left out of the initial draft while others argued the group should continue with plans to merge IMHP into the draft. Documentation and Implementation Milestones The group needs a specification which can be used to implement test systems (would like the specification before Christmas). Charlie will continue work as the document editor. Interim Meeting An Interim meeting of the Mobile IP Working Group was proposed for January at Xerox PARC. It was suggested that implementors and specification writers convene for two days. Attendees Kannan Alagappan kannan@dsmail.enet.dec.com Kenneth Albanese albanese@icp.net Nick Alfano alfano@mpr.ca Stephen Batsell batsell@itd.nrl.navy.mil Tom Benkart teb@acc.com Mark Beyer beyer_mark@tandem.com Ram Bhide ram@nat.com Steven Blair sblair@us.dell.com Jon Boone boone@psc.edu Monroe Bridges monroe@cup.hp.com Glen Cairns cairns@mprgate.mpr.ca Ken Carlberg Carlberg@cseic.saic.com Lida Carrier lida@apple.com Bill Cash cash@bangote.compaq.com Bilal Chinoy bac@sdsc.edu Frank Ciotti frankc@telxon.com David Clark ddc@lcs.mit.edu Thomas Coradetti tomc@digibd.com Stephen Deering deering@parc.xerox.com Thomas Dimitri tommyd@microsoft.com Waychi Doo wcd@berlioz.nsc.com Avri Doria avri@locus.com Robert Downs bdowns@combinet.com Pierre Dupont dupont@mdd.comm.mot.com Julio Escobar jescobar@bbn.com Craig Fox craig@ftp.com Richard Fox rfox@metricom.com John Garrett jwg@garage.att.com Robert Gilligan Bob.Gilligan@Eng.Sun.Com Ramesh Govindan rxg@thumper.bellcore.com Darren Griffiths dag@ossi.com Robert Grow bob@xlnt.com Regina Hain rrosales@bbn.com Jari Hamalainen jah@rctre.nokia.com Marc Hasson marc@mentat.com Cornelius Healy con@icp.net Juha Heinanen juha.heinanen@datanet.tele.fi Kathryn Hill khill@newbridge.com Robert Hinden hinden@eng.sun.com Kevin Jackson kjackson@concord.com David Jacobson dnjake@vnet.ibm.com B.V. Jagadeesh bvj@novell.com David Johnson dbj@cs.cmu.edu Timo Jokiaho timo.jokiaho@ntc.nokia.com Rick Jones raj@cup.hp.com Elizabeth Kaufman kaufman@biomded.med.yale.edu Byonghak Kim bhkim@cosmos.kaist.ac.kr Mark Knopper mak@merit.edu Tony Li tli@cisco.com Tracy Mallory tracym@3com.com Wayne McDilda wayne@dir.texas.gov Marjo Mercado marjo@cup.hp.com Greg Minshall minshall@wc.novell.com William Miskovetz misko@cisco.com Randy Miyazaki randy@lantron.com Robert Moose rmoose@gateway.mitre.org Sandra Murphy murphy@tis.com Andrew Myles andrew@mpce.mg.edu.au Erik Nordmark nordmark@eng.sun.com Masataka Ohta mohta@cc.titech.ac.jp Todd Palgut todd@nei.com Steve Parker sparker@ossi.com Ismat Pasha ipasha@icm1.icp.net John Penners jpenners@advtech.uswest.com Charles Perkins perk@watson.ibm.com Wayne Peters waynep@telxon.com Ram Ramanathan ramanath@bbn.com Jim Rees Jim.Rees@umich.edu Yakov Rekhter yakov@watson.ibm.com Mike Ritter mwritter@applelink.apple.com Benny Rodrig brodrig@rnd-gate.rad.co.il Greg Ruth gruth@gte.com Richard Schmalgemeier rgs@merit.edu Martin Schulman schulman@smtp.sprint.com Dallas Scott scott@fluky.mitre.org Isil Sebuktekin isil@nevin.bellcore.com Michael See mikesee@vnet.ibm.com Satya Sharma ssharma@chang.austin.ibm.com William Simpson Bill.Simpson@um.cc.umich.edu Henry Sinnreich hsinnreich@mcimail.com James Solomon solomon@comm.mot.com Michael St. Johns stjohns@arpa.mil Martha Steenstrup msteenst@bbn.com Robert Stevens robs@join.com David Stine dsa@cisco.com John Tavs tavs@vnet.ibm.com Fumio Teraoka tera@csl.sony.co.jp Susan Thomson set@bellcore.com Akihiro Tominaga tomy@sfc.wide.ad.jp Paul Traina pst@cisco.com Hoe Trinh htrinh@vnet.ibm.com Keisuke Uehara kei@cs.uec.ac.jp John Veizades veizades@ftp.com Gerry White gerry@lancity.com Steve Willens steve@livingston.com Bradley Wilson wilson@ftp.com David Woodgate David.Woodgate@its.csiro.au Richard Woundy rwoundy@vnet.ibm.com Honda Wu honda@nat.com Jean Yao yao@cup.hp.com Weiping Zhao zhao@nacsis.ac.jp