[IMR] IMR87-01.TXT Westine [Page 1]  ~ JANUARY 1987 INTERNET MONTHLY REPORTS ------------------------ The purpose of these reports is to communicate to the Internet Research Group the accomplishments, milestones reached, or problems discovered by the task forces and contractors in the ARPA Internet Research Program. This report is for research use only, and is not for public distri- bution. Each task force and contractor is expected to submit a 1/2 page report on the first business day of the month describing the previous month's activities. These reports should be submitted via ARPANET mail to Westine@ISI.EDU. Reports are requested from BBN, ISI, LL, MIT-LCS, NTA, SRI, UCL, and UDEL. Other groups are invited to report newsworthy events or issues. BBN LABORATORIES AND BBN COMMUNICATIONS CORPORATION --------------------------------------------------- WIDEBAND NETWORK NETBLT experiments performed over the Wideband Network this month yielded substantially improved results as compared to previous experiments. Observed packet losses were generally 1% or less. The correction of hardware problems in ISI's ESI is believed to be Westine [Page 1]  Internet Monthly Report January 1987 a contributing factor to the improved results. BSAT software has been built to run under Chrysalis Release 2.3.1, using a new C compiler which generates more efficient code than the old compiler. The new BSAT code has run successfully on the Wide- band Net and will be released soon. Packets were successfully transmitted between two BSMIs in real time using an 8 Mbps interface clock speed. For this test the BSMIs' on-board 68020 processors ran code emulating the functional- ity of the BIO hardware currently used in the BSATs. The BSMIs were driven by Butterfly processor nodes running a modified version of standard synchronous I/O software. This BSMI/PNC software will be integrated into the BSAT software in the initial BSMI version of the BSATs. A number of improvements and corrections were made to Wideband Net- work earth stations. Antenna repointing was performed at Lincoln and SRI. New de-icing equipment was installed at BBN and CMU and has so far exhibited improved reliability. Corrections were made to Western Union's control of the HPA subsystems at a number of sites. Work performed at Lincoln appears to have eliminated the periodic connectivity outages that had been occurring there. VAX UNIX NETWORKING In the month of January, design, implementation and testing of the Inter-Agent protocol for passing group membership information between Multicast Agents was begun. Karen Lam has been working with Steve Deering of Stanford University to write a specification for the protocol. At the End-to-End Task Force meeting, held 12 - 13 January in Los Angeles, several parties expressed an interest in receiving the 4.3bsd Unix multicast implementation. SATNET During January, the SATNET was mostly stable. However on January 14, the SATNET experienced some hardware problems with the PSP ter- minals. Problems with some of the channel modules caused Tanum to broadcast clean carrier on channel 1 and Fucino to broadcast clean carrier on channel 0. Fucino had spare equipment which was installed but Tanum is now only using channel 0. Channel 0 has remained healthy and work was done to adjust transmit and receive levels. We continued to work on obtaining spare Linka- bit modems and they are now about to be tested at Comsat. Westine [Page 2]  Internet Monthly Report January 1987 GATEWAYS On January 20 the UCL Butterfly Gateway was reinstalled on Satnet replacing the PDP-11 gateway. A Butterfly Gateway was also installed at DCEC on January 28 replacing the old PDP-11. DCEC was the last PDP-11 Gateway on Satnet. Now all the the gateways (CSS, DCEC, UCL, RSRE, CNUCE, and NTA) are Butterfly Gateways. Since about January 23 the Arpanet has been severely congested. This is causing the gateways (core and non-core) to break their routing connections (EGP, GGP, SPF) resulting in much reduced internet service over the Arpanet. We are currently looking at how we can tune and modify the routing protocols to provide better ser- vice when the Arpanet is congested. Bob Hinden ISI --- Internet Concepts Project ISI hosted the END TO END Services Task Force, the Satnet Measurement Group, the SATNET and Infrastructure Group, and the International Collaboration Board meetings the week of 12-16 January. Greg Finn is finishing an ISI report on his research into a simple routing scheme for large scale internetworks. The draft will be finished by the end of the month. Two RFCs were published: RFC 994: ANSI X3S3.3, "Final Text of DIS 8423, Protocol for Providing the Connectionless-mode Network Service" RFC 995: ANSI X3S3.3, "End System to Intermediate System Routing Exchange Protocol for Use in Conjunction with ISO 8473. Multimedia Conferencing Project Our teleconferences with packet video have been limited to two sites because we have only two sites equipped. We are working to develop additional sites to allow research in management of multi-site conferences. Since other groups have expressed interest in packet video, we are investigating the possibility of adapting a commercial video codec to the packet network instead of building more of our own. This would provide the Westine [Page 3]  Internet Monthly Report January 1987 option to extend the system beyond four sites, which would not be practical with our experimental codec. The commercial ven- dors have advanced the technology considerably since we began work on our codec, but the problem that led us to seek our own solution still exists: the commercial systems do not tolerate packet loss well. It's clear from the multimedia teleconferences we have con- ducted that while all the media work together to make the conference effective, voice is the most crucial to carry hard information. More work is required to make the audio system fully satisfactory. We are setting up a permanent teleconfer- ence room at ISI where the audio system can be more carefully tuned. A special table is being constructed to allow up to six people to easily see the shared-workspace media on two Sun screens without blocking the video camera and monitor. Having the room always available will also promote use of the system to give us more experience. Steve Casner visited Pictel Corporation in Boston, MA, 27-28 Jan, to discuss Video Codecs. Steve Casner Brian Hung was able to send multimedia messages containing bitmap data from his IBM-PC AT to a Sun workstation running the Diamond multimedia system and have the messages displayed on the screen. The problem he has been having was due to a misspelled word in the multimedia messages he was generating. Brian plans to integrate the current document scanning and multimedia message software into one program and to look into the feasibility of adding text capability into his multimedia message software package. Brian Hung NSFNET Project Bob Braden and Annette Deschon hosted a meeting of the End- to-End protocols Task Force at ISI, Jan 12th and 13th. In this connection, Annette prepared a critique of the draft RFC on the SUN data representation standard XDR. Bob Braden also gave an ISI seminar on reliable multicasting, reviewing the work of Chang and Maxemchuk. Work continued on the RFC985 revision and on the background file transfer problem. Westine [Page 4]  Internet Monthly Report January 1987 Bob Braden made a presentation to a session at the Uniforum meeting in Washington, DC, 20-23 Jan, on "The Internet Archi- tecture -- Present and Future Internet. Bob Braden Supercomputer and Workstation Communication Project 1. Finished a comparison of how mathematical equations are entered and represented in various text processing systems. 2. Wrote "Issues in Defining an Equations Representation Stan- dard" which should be distributed as an RFC shortly. 3. Began a survey of network accessibility of the NSF supercom- puter sites. Most of them can now be reached over the Inter- net via Telnet, but not all of these have a working FTP capa- bility (yet). Over the next few months, Alan plans to do a series of tests with each site where he will try to login via Telnet, transfer a test Fortran program created on my worksta- tion via FTP, run the program on the supercomputer, and get the results back on his workstation (perhaps with some graphi- cal display). 4. Experimented more with X-windows, which he hopes to be able to use eventually to interact with supercomputers from his works- tation. Alan Katz MIT-LCS ------- No internet related progress to report for January 1987. Lixia Zhang NTA & NDRE ---------- No report received. Westine [Page 5]  Internet Monthly Report January 1987 SRI --- No report received. UCL --- Jon Crowcroft and Bruce Wilford (Head of the UCL Service Project for Network Interconnection) collectively attended the SATNET Meas- urements Task Force meeting, the End to End Protocols Task Force meeting and the SATNET/Infrastructure and ICB Meetings at ISI. Steve Easterbrook has started work on the DARPA Autonomous System Management Project at UCL. Currently he is working on enhancing PEPY, the ISODE Presentation layer ASN parser. Jon Crowcroft UNIVERSITY OF DELAWARE ---------------------- 1. I reviewed the comments received on my proposed authentication scheme, revised the proposal and submitted it for considera- tion as an RFC. I also documented the UDP-based network statistics server for the NSFNET fuzzballs and submitted it for consideration as an RFC. 2. I completed a preliminary draft proposal for the Dissimilar Gateway Protocol and submitted it to the task forces for review. This proposal is presently undergoing substantial development and revision. 3. The fuzzball software configurations at Linkabit, Ford and the University of Delaware were amended to support a massive swap in network numbers and related information. This turned out to be much harder than imagined, since many of the name and address changes required TSRs and HAFs. There were some prob- lems with blizzards, too. The changes have not been completely implemented yet. 4. The problems reported last month with the X.25 link between the NSFNET gateway at CMU and its PSN seem to have been resolved. The latest scatter diagrams show delays reduced by a factor of four and throughputs increased by a factor of seven. Westine [Page 6]  Internet Monthly Report January 1987 Now the bad news. The EGP gateway system has been in shambles for the last several days. Peer paths are bobbing up and down like corks at all the gateways which can be monitored. Hans- Werner Braun at the University of Michigan and I had to resort to tunneling ARPANET port-expander type addresses through the NSF swamps to get even this information. 5. I collected connectivity data for all the known NSFNET back- bone, regional and campus nets now operating and constructed a data base following the model proposed for DGP. The truly awe- some topology was used as the basis for the examples included in an appendix to the proposal. 6. The fires were kept warm in the fuzzball smelter. USECOM Patch Barracks (Stuttgart) sputtered back to life on MILNET and fur- nished interesting end-end delay measurements which Mike Min- nich here is stuffing into some interesting mathematical models suggested by Van Jacobsen of LBL. I made changes in the fuzzware to improve fairness to transit traffic in the face of massive abuse from local Ethernets. Dave Mills TASK FORCE REPORTS ------------------ APPLICATIONS -- USER INTERFACE The instance of the applications task force instantiated in January 1985, and chaired by Bob Thomas, no longer exists. Indeed, it ceased to function about 18 months ago and was for- mally disbanded by the IAB four months ago. Consider this the (belated) notice of its demise. However, I am currently in the process of forming a User Interface Task Force (UITF) for the DSAB (Distributed Systems Activities Board), which for the moment will double as the applications task force of the IAB -- by joint approval of the IAB and DSAB. The UITF will investigate the requirements and make constructive proposals for improved user interfaces to distributed computing environments. Features of distributed environments that distinguish them from non-distributed environments and impact the design and implementation of user interfaces include: - additional constraints on delay and bandwidth between applications and users; Westine [Page 7]  Internet Monthly Report January 1987 - increased heterogeneity -- of both hardware and software; - new opportunities for loosely-coupled parallel processing; and - the need to support collaboration between geographically distributed participants. The task force will address these and related issues primarily with respect to requirements definition and implementation strategies -- although aesthetic issues will not be ignored. Initial efforts will be focused in two inter-related problem areas: user interface architecture and multimedia conferenc- ing. Graphics packages, window systems, and user interface manage- ment systems reflect three major trends in user interface design. Until recently, however, little attention has been paid to the impact of distribution on or the inter- relationships between these three types of systems. The task force will endeavor to address these issues, to develop a reference model for user interfaces to advanced distributed computing environments, and to propose some alternative imple- mentation architectures. Multimedia conferencing is the outgrowth of one of the major value-added features (as opposed to disadvantages) of distri- buted computing, namely, the ability to lend computer support to the cooperative work of geographically distributed partici- pants. The task force will investigate existing multimedia conferencing systems, attempt to distill common principles, and apply those principles in the context of the unified user interface architecture to be developed in parallel. If you are interested in participating in this task force, and haven't already been invited individually, please let me know. Keith Lantz AUTONOMOUS NETWORKS The Autonomous Networks Task Force has a meeting planned for March 20-21 in Palo Alto. Deborah Estrin Westine [Page 8]  Internet Monthly Report January 1987 END-TO-END SERVICES HOST GROUP MULTICASTING The continuing collaboration of Steve Deering at Stanford and Karen Lam at BBN will soon lead to distributable implementa- tions of the Host Group Multicasting (HGM) facility (RFC966, RFC988). Karen has successfully modified the IP code of 4.3BSD for full host participation in HGM, and she now has experimental multicast agent code for 4.3BSD. A VAX at BBN running this code is collaborating with a V-based multicast agent at Stanford. Eric Cooper announced that Karen's IP code with with its HGM capability will be used as the IP code for MACH at CMU. He plans to make use of the HGM facility to provide Internet- transparent multicasting. There have been healthy and useful interactions with various R&D groups interested in multicasting -- the NETBIOS vendor group, the SURAN effort (Rockwell), and Ken Birman of Cornell. One result has been some minor changes in the protocol and some implementation enhancements to improve the robustness and crash-recovery ability of HGM. TRANSACTION PROTOCOLS Dave Cheriton has produced an extensive document on the VMTP protocol, which will be submitted as an RFC after review by the Task Force. A couple of students have implemented a VMTP subset in the BSD kernel, and Dave hopes to have this avail- able for limited distribution later this Spring. The Task Force is also evaluating the TTP protocol being developed for SURAN, and would like to consider the REX proto- col of the UK ANSA project. DATA STRUCTURING The group discussed the draft RFC from SUN proposing XDR as a standard data representation. In paricular, the comparison was between XDR and X.409. Possible deficiencies which were noted in XDR as a wide-use Internet standard were: its use of implicit types (preventing type checking or application- independent decoding), its lack of a bit-string type, its ina- bility to talk about data items shorter than 32-bits, null values, or structured constants, and its inability to specify byte-swapping. Westine [Page 9]  Internet Monthly Report January 1987 A data structuring language has two possible functions -- for implementation and for description. It seemed that XDR is not strong as a description language, and would perhaps benefit from a better abstract syntax. NETWORK FILE SYSTEMS Rusty Sandberg of SUN attended the meeting to discuss a draft RFC on SUN's NFS. The agenda for this discussion was set by a list of a dozen NFS problem areas, raised by many contributors to a recent blast on the TCP-IP mailing list. Rusty described changes SUN is proposing to meet some of these objections (e.g., support for record-oriented I/O, and fixes for idempo- tency failures), and explained the rationale behind some of the controversial points. The Task Force has a major concern about the hidden dependen- cies of SUN's NFS upon SUN's particular RPC protocol in a lower layer. This seems to be a very undesirable departure from reasonable protocol layering; an acceptable Internet standard network file system should have a well-defined ser- vice interface to the underlying protocols, and be implement- able independently of lower layer implementations. INTERNET NETBIOS Avnish Aggarwal of Excelan, chairman of the NETBIOS Technical Task Force, and Paul Mockapetris of ISI, attended the meeting to discuss plans for implementing NETBIOS over the Internet. The primary issue is how best to handle the NETBIOS require- ment for multicasting across the Internet. Two approaches have been suggested -- the HGM facility (RFC966/988), and the addition of a dynamic update capability to the Internet name servers. The HGM scheme seems to lead to the simpler imple- mentation, both conceptually and in detail; however, there are serious questions raised by the delay characteristics of the Internet. NETBIOS applications have been designed assuming packet delays characteristic of a LAN, and might need to be redesigned if those delays rose to 10's of seconds. Bob Braden Westine [Page 10]  Internet Monthly Report January 1987 INTERNET ARCHITECTURE 1. Two documents were circulated for review and comment, one on an authentication scheme designed for EGP and similar protocol models, and the other on a proposed hierarchical gateway model. 2. INARC related issues were discussed on several lists, includ- ing those related to TCP retransmission strategies (tcp-ip) and host/gateway monitoring (gwmon). 3. New documents were received from ANSII on (a) connectionless transport service, (b) connectionless transport protocol and (c) operation of ISO IP over HDLC links. Documents (a) and (b) define what might be called ISO UDP, while document (c) may shake the cobwebs from the important isssue of how to get multi-vendor gateway systems to play over serial lines. Dave Mills INTERNET ENGINEERING No report received. PRIVACY No report received. ROBUSTNESS AND SURVIVABILITY No report received. SCIENTIFIC COMPUTING The White Paper "Networking Requirements for Scientific Research" was completed and submitted to the FCCSET Network Subcommittee. It will be discussed at the upcoming Workshop on Computer Networks. The report is available from RIACS upon request. Barry Leiner Westine [Page 11]  Internet Monthly Report January 1987 SECURITY No report received. TACTICAL INTERNET No report received. TESTING AND EVALUATION No report received. Westine [Page 12]