Applications Area meeting Dallas, March 20, 2006 Notes by Paul Hoffman Follow the slides from This is just a summary Brief intro of Lisa Dusseault as Scott Hollenbeck's replacement Some folks couldn't get into the hotel, so one part was moved down in the agenda WG status EDIINT closed GEOPRIV and SIMPLE moved to RAI Directory Directorate meeting cancelled LDAP-aware people will get together informally Who is my AD? List is split between Ted Hardie and Lisa Ted and Lisa might trade stuff later as things progress WGs and BOFs this week EAI is now a WG meeting Looking for a co-chair Want to have experimental phase before standards track DIX Lots of discussion so far Dave Crocker asks what problem are they solving. Potentially interesting cross-area BOFs HOAKEY (which takes in the preauth work) P2P-SIP CRISP Internet Routing Registry (IRR) proposal Presented by Tomoya Yoshida, JPNIC IRR team chair Agenda Problems with IRR New approach Next steps Current issues Stores internet routing registry Many organizations run RIR Based on RPSL Main use is make their BGP filtering useful Cannot know where the various IRRs are Mirroring issue Not full mesh, so there are many mirroring paths Freshness issue Mirroring causes timeliness issues depending on the links Some mirrors refresh 2 hours, some one day There is no standard refresh time Query limitations Size of responses will depend on mirroring Discussion is happening in CRISP WG IRR is not based on a real root Uses same structure as SBGP PKI infrastructure CRISP is about search, not registration It is hard to acquire accurate information about the topology Current proposal Make a new IRR framework work wiht SBGP PKI Many hard questions How to operate in the future? Who should run the IRR How to keep it fresh? Hierarchical? George Michaelson (CRISP WG co-chair) WG consideration is operation is non-hierarchical even though the data is Is it appropriate for the work to be done here? Referrals will scale even though mirroring will not Geoff Huston, APNIC The data differ between routing registies Mirroring causes overlaps of inconsitent data What do you do with these conflicts, given that they are purposeful? Ted Asks if there are current data structures that will reflect the policies. Asks if the current proposal removes the context or uses it George Difference between what operations want to declare publicly vs. what they want to operate with privately Geoff It's not incorrect information, it is correct in a particular context CRISP assumes a hierarchy, so maybe CRISP is the wrong Andrei Robachevsky, RIPE NCC Routing registries is contextual, so information is contextual Certification can help solve the problem without creating a global registry Allison Mankin Difference between policy and announcements IRIS has flexibility to create right data objects Maybe not use RPSL format Talk to SIDR and think more creatively Ted JPNIC use cases may be different May be a way of splitting the problem of representation vs. correctness Is there a similar take that could be used by SIDR Geoff SIDR is in the last throes of chartering Maybe bring it to that context George If you wrote well-formed XML for routing info, work could be adoptied, as long as it is divorced from operations Still wants a scheme for searching across databases John Klensin on FTP internationalization No slides RFC 2640 claims to do FTP i18n Only about directories and files, not content This brings up FTP binary vs. text transfers Mostly can be ignored now other than line endings Is it time to invent a third type? Doesn't want to do it himself Are others interested? Maybe form a small team George Michealson asks why a third option is needed? Maybe just deprecate A mode CRLF misuse is widespread Unless we deprecate type A and put it on the OS, John sees a need Dave Crocker says that if there is a strong consumer base, we should respond There are filesystems that handle UCS2, UCS4, UTF8, etc. Wants a new type for moving these files around Need to invent an NVT-type format based on Unicode FTP is the most obvious first place IETF is the right place because the people doing it need to understand the protocol implications Mike Andrews thinks we need to go this way Distributed Multimodal Synchronization Protocol Presented by Chris Cross Also will be discussed in the RAI area Problem: distribution of multimodal web applications Other IETF work MSCP LRDP MRCPv2 Widex Why you might need a distributed system such as a thin client Some small devices may not be able to solve all problems Good example: speech recognition Tradeoff is bandwidth Other examples are CPU intensive (handwriting recognition) Buiding blocks Way a user interacts with an app Model-view-controller design pattern is central View-independent Event-based modality sync: protocol Modalities are views (GUI, speech, pen) Compound browsers have two or more modalities View-independent model enables a centralized model Need to be able to synchronize interactions between modalities Can be handled in the network with a protocol Details of DMSP Many message types Both data and control messages needed Events are the meat of the sync problem Wants people to work with Eric Berger agrees that this is a good tool and hopes others will review the documents Flows out of VoiceXML but can be used in a much wider context Vlad Stirbu Many things here were discussed in Widex A lot of similarities between the two efforts WG charter doesn't limit the modalities, only has to be in XML John Klensin points out acronym overlap Once you pick a language, it becomes the least common denominator Open mic DIX discussion Dave Crocker asked what problem is DIX solving Lisa says that the problem to be solved in the charter Having home be able to say who you are BOF has one proposal so far, but very open Paul Hoffman asked about the overlap with Security Area Bob Morgan says this might be used in a low-security (no-security) context while also fitting in with "real" security protocols Chris Newman says that Liberty Alliance is working on the three-party model where the client is only a web browser Bob Morgan corrects that it is not just web browsers, it also covers web services Hannes Tschofenig says that SIP is working on SAML in SIP They would like more help Finished up at 10:45