[IMR] IMR86-09.TXT SEPTEMBER 1986 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 distribution. 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, LINKABIT, ISI, LL, MIT-LCS, NTA, SRI, and UCL. Other groups are invited to report newsworthy events or issues. BBN LABORATORIES AND BBN COMMUNICATIONS CORPORATION --------------------------------------------------- SATNET The SATNET continued to be stable. As before, we are holding off on correcting channel 1 problems until Linkabit delivers some spare modems. Since channel 0 performance was fine, the SATNET provided uninterrupted service. Last month's release of a new version of the Butterfly gateway software seems to have greatly improved the stability of the gateways at CSS, NTARE, CNUCE and RSRE. However, there are still some problems left to be solved. Until then, we are continuing to use PDP/11 gateways at DCEC and UCL. Claudio Topolcic attended the SATNET Measurement Taskforce meeting at SHAPE Technical Center, The Hague, Netherlands on 9/24/86. He also visited UCL, RSRE and NTARE to discuss the ongoing measurement work with taskforce members who could not attend the meeting. Alex McKenzie, Bob Hinden and Claudio Topolcic attended the SATNET/ICB meeting that was held at SHAPE Technical Center on 9/25/86. WIDEBAND NETWORK This month's efforts focused on investigating various Wideband satellite channel problems and completing the development of features to be included in the next BSAT software release. Network-wide difficulties with the reception of satellite channel bursts were tracked down and found to be a frequency problem in the upconverter at the ISI site. In addition, a slope in the noise floor of the signal received at the BBN site was determined to be caused by a faulty RF bandpass filter at BBN.  Automatic monitoring of the power levels of bursts received on the Wideband channel from selected sites was initiated at BBN in the latter half of the month. The aim of this monitoring is to characterize any changes in inter-site power level differences over time, and to determine the source of such variations. Large enough inter-site power level differences in a TDMA network such as the Wideband Net can result in a degradation of the effective satellite channel performance. Relative power level changes that occur on a daily or monthly cycle may be indicative of antenna pointing problems at the sites in question. Preliminary results have shown the power level difference between the bursts received from the ISI and BBN sites to vary by 3.5 db. The software for BSAT Release 2.1 has been frozen and is undergoing final testing. A Butterfly multiprocessor for use in support of the BSAT software development and testing was delivered at the end of the month. VAX UNIX TCP/IP In the month of September, the design and implementation of the "Multicast Agent" for 4.3 BSD Unix was begun. The multicast agent will be implemented partially inside the Unix kernel, and partially as a user-level daemon. As the formal specification for the agent has not yet been finalized, continuing discussion has gone on between Steve Deering of Stanford University, and Karen Lam. The plan is for the two of them to write the final specification together. Testing of the IGMP implementation in the 4.3 BSD kernel also continued during the month of September. GATEWAYS The new software installed in the Satnet Butterfly Gateways is working well. We have begun discussing with UCL putting the UCL Butterfly Gateway back on Satnet. In order to provide better service in light of the current ARPANET congestion problems, we have begun changing the Butterfly Gateways on the Wideband network to advertize a lower cost for the Wideband network than the ARPANET Butterfly Gateways advertize for the ARPANET. This will cause LAN-LAN traffic (for LAN's which are connected to the Wideband network) to go over the Wideband network instead of the ARPANET. This traffic should see reduced and more consistant delays and of course higher throughput. The majority of the LSI-11 gateways (Mailbridges and EGP servers included) are now running the 300 network software. We are working to install this software in the remaining LSI-11 gateways (Packet Radio and ARPANET HDH). Bob Hinden  ISI --- General Bob Braden, Annette DeSchon and Paul Mockapetris visited University College London (UCL) in London, England, Sep 1-5, to discuss current research topics in the area of computer networks. During their visit they met with Jon Crowcroft, Ben Bacarisse, and Mark Riddoch of the UCL Department of Computer Science. They discussed the Sequential Exchange Protocol (SEP) and a companion Remote Procedure Call (RPC) system which are being developed at UCL for use in the ADMIRAL project. SEP is a transaction protocol that is designed to support sequential exchanges of potentially large messages. It is the basic for a RPC system that has been used to implement experimenal applications including file server access and workstation mail support. Also discussed were plans for a joint UCL/ISI test using an implementation of SEP Sequential Exchange Protocol which runs on a Sun 3 workstation. This test will make use of the SATNET link between UCL and the ARPANET and will provide data on the performance of SEP over the Internet. On September 3-4, Annette and Bob attended a meeting of the DARPA End-to-End Task Force, and on September 5th they attended the International One Day Workshop on Protocols for Distributed Systems. Paul Mockapetris presented a paper titled "Implementation of the Domain Name System" at the SIGOPS conference in Amsterdam, September 8-11. Jon Postel participated in the IEEE Workshop on Computer Communications, in Warner Springs, CA. September 24-26. Jon Postel and Bob Braden attended the NSF-NET Advisory meeting in Washington, D.C., September 20. Jon Postel and Ann Westine Internet Concepts Project See the General section above. Multimedia Conferencing Project Steve Casner attended a demo of the SRI Multimedia Conferencing system on Monday at NOSC. And see the General section above. Steve Casner Supercomputer and Workstation Communication Project See the General section above. Computer Center No internet-related progress to report. MIT-LCS ------- August & September Report from MIT-LCS 1. A new version of the NETBLT protocol specification has been submitted for NIC RFC publication. The revision was based on extensive testing of NETBLT's performance over long-delay, high-bandwidth satellite channels. Most of the changes are related to the computation and use of data timers. 2. A new version of the Pcmail specification has also been submitted. The revision was based on discussions and comments from a variety of sources, as well as further research into the design of interactive Pcmail clients and the use of client machines other than IBM PCs. One of the major changes is the addition of the bboard support. 3. Lixia Zhang presented a paper, "Why TCP Timers Don't Work Well", at the SIGCOMM'86. She also attended the TCP-IP Vendor's Workshop in Monterey, 25-27 August. 4. Recently we have observed very poor ARPANET performance. Intolerable network delays and frequent connection closes make the network virtually unusable during daytime. Lixia Zhang NTA & NDRE ---------- No report received. SRI --- SRI successfully demonstrated the multimedia conferencing system (EMCE) at NOSC on September 29. This conferencing system prototype is implemented on SUN3/160C workstations with digitized voice input from Speech Periperial Processors (SPPs). The main "content" (subject) of the conference is a Naval situation map showing friendly, neutral, and opposing forces derived from real Naval messages. These messages can be generated and incorporated in real time from a simulator (Interim Battle Group Tactical Trainer). EMCE allows a dialogue composed of digital voice and grpahic pointers to occur between multiple participants viewing the same contents. Local control of the viewing of the contents (ie, resizing the main conference window) can be accomplished while maintaining the pointing references (the conference pointer will circle the same object in all conferees' windows). The conference floor control is distributed, using a "talker sense" algorithm (if your receiver is active you can't talk). Limited feedback from conferees that do not have the floor is accomodated giving a more natural interchange (interruptions, agreements, etc..) and maximizing channel utilizatiion for fixed allocation channels (circuits). The standard Sun window interface (SunView) has been augmented by a SRI-developed special display package called PLView written primarily in prolog. This package has been integrated with the Navy-sponsored software to give very flexible data representation and access. Natural language and reasoning modules can be easily added. The content of the conference is replicated in all workstations (shared workspace) for robustness and rapid response. Additional text and image information related to the main contents (the situation map) can be displayed in a second window. Future plans include allowing the conference pointer to be in this window, provide cross window references (show a picture of this ship) and various manipulations (image magnification). SRI has completed the recoding and testing of a new MPM software module for multimedia mail to run on Sun UNIX version 2.2 and 3.0. The SRI MMM system and BBN'S Diamond can use this module for MM mail transport. The new version fixes several bugs (thanks to Sun!!) that would not allow operation on diskless clients. The code has been rewritten to made it much faster and efficient. Domain addresses of the type now appearing in the internet host tables can be handled somewhat intelligently (this is still short of using a full domain name server). MPM can now be forced to cycle (deliver/receive mail) faster at times other than one hour. Many flavors of make files are available for various Sun host configurations (diskless-diskful, YP-nonYP, 2.2/3.0 mix, Sun2/Sun3, ...) Earl Craighill UCL --- Report from UCL for September 1. New PROMS and software were installed in the UCL Butterfly Gateway at the beginning of the month. Since then, we have been directing a proportion of our service traffic through it as a test. For the moment, we are still using the PDP-11 gateway as the direct link to Goonhilly; the Butterfly connects to Goonhilly via RSRE. Many of the earlier faults with the Butterfly appear to be cured, but a problem remains with the exchange of routing information - this causes the gateway to become isolated approximately once a day, necessitating a restart. 2. We have been conducting some multimedia message tests using the BBN Diamond system recently installed at UCL. Large multimedia messages, e.g. longer than 13K bytes, have been successfully sent from UCL to the Diamond system at BBN. An initial problem was caused by a buffering restriction in a local gateway on the default route through UCL - the large buffers necessary for multimedia traffic tended to kill the performance of the other services through that gateway. Instead, we persuaded the Diamond system that it was really on a separate UCL network so that it could talk directly to the UCL Butterfly Gateway. From there, traffic went via RSRE to Goonhilly. The current problem is that the remote end appears to receive multiple copies of each message sent from UCL. 3. Work has started with Bob Braden's benchmark for presentation levels, looking at the Marshall Rose ASN.1 system in his ISODE and the UCL Courier-like interface language. Initial measurements agree with Eric Cooper's analysis that the time-critical part of performance is allocating buffers for the type conversions. The UCL ESP protocol service interface allows conversion to be done from buffers straight to the protocol. This results in a speed up of around 50% in stub code for simple examples. The ROS/ISODE equivalent and the more complex example are being worked on. 4. At the beginning of the month, UCL hosted a meeting of the End-to-End Task Force and also an International Workshop on Protocols for Distributed Systems. Topics at the latter included Integrated Services on LANs, MIT's NETBLT, the MACH/Camelot system, the Newcastle Connection, Internet multicasting, the UK ANSA's Remote Execution Protocol, and Stanford's VMTP. 5. Claudio Topolcic (BBN) visited UCL in the middle of the month and had discussions with Peter Lloyd about SATNET measurements. Peter Kirstein and Jon Crowcroft attended the SATNET & Infrastructure and ICB meetings at STC, Holland. Peter Lloyd UDEL ---- 1. Enough of my gear from Linkabit has been moved to the University of Delaware to bring up a fuzzball on the UDel Etherswamp. It's first duty was to facilitate a few fixes which improved gateway performance between that swamp and the ARPANET. 2. A good deal of muscle went into moving and reconfigurating equipment at Linkabit in order to sustain ARPANET connectivity for the NSFNET crowd and improve maintainability. Linkabit will continue to operate that swamp (with my help) for mail distribution and for support of other projects. 3. Effort continued on the design and simulation of routing algorithms for MSS. I presented preliminary results of this study at the at the SURAN Implementors meeting at BBN. 4. The NSFNET Backbone traffic continues to escalate, tripling during the month to about a megapacket per week via DCN-GATEWAY, presently the only ARPANET valve feeding this plumbing. A dedicated gateway now being installed at CMU is to pick up most of this traffic and relieve DCN-GATEWAY and a backup gateway at Cornell. The new Unix gateway daemon mentioned last month is functioning well in the Cornell gateway and even survived panic for a week when DCN-GATEWAY was isolated due to broken lines and IMPs. While the conglomerated NSF swamp is presently functioning well, many issues involving stability, metric compatibility and core-hop ambiguity remain to be resolved. 5. I implemented, installed and tested a new Network Time Protocol (NTP) daemon for the fuzzball. The new daemon operates in symmetric mode with multiple peers, as suggested in RFC958, and employs some of the fancy filtering and smoothing algorithms described in RFC956. Preliminary results suggest clock synchronization to within a few tens of milliseconds should be possible via typical Internet paths. Eventually, this daemon will be installed in the various isolated fuzzball swamps that do not have local radio clocks. Dave Mills  TASK FORCE REPORTS ------------------ APPLICATIONS No report received. END-TO-END SERVICES No progress to report this month. INTERNET ARCHITECTURE 1. Preparations continue for the upcoming meeting at SRI. I have invited Scott Brim of Cornell to discuss NSFNET routing issues and Doug Comer of Purdue to discuss the CYPRESS project and its impact on the Internet. Phill Gross and I worked out a schedule for the joint INENG/INARC meeting day, which will involve presentations from both communities on issues of common interest. 2. A special-interest group has formed to study issues involved with campus networks, in particular the state and regional networks associated with the NSF community. The participants, including some INARC and INENG members, as well as several gurus and wizards from various organizations, have set up a mail list on which to babble. Dave Mills INTERNET ENGINEERING No report received. INTEROPERABILITY Deborah L. Estrin is replacing Robert Cole as the chairman of the Interoperability Task Force. The next activity will be to review the current version of the "Research Issues in Autonomous Systems" draft, and to establish a more concrete agenda for how to proceed from there. If anyone is interested in participating in the Task Force please contact Deborah at "estrin@USC-CSE.USC.EDU". Deborah L. Estrin PRIVACY The Privacy Task Force held a successful 2-day meeting on 11-12 September at UCL, hosted by Steve Wilbur. Attending members were: Steve Kent, John Linn, Dan Nessett, Mike Padlipsky, Rob Shirey, and Steve Wilbur, with guests Danny Cohen, Charles Fox, Jim Hopkins, Steve Kille, and John Laws. We discussed a set of changes to be made to the draft RFC on electronic mail encipherment and authentication procedures. An option allowing selective application of encryption will be provided, and procedures involved in manual dissemination of interchange keys will be discussed. We expect that a revised RFC will be reviewed by the membership and distributed to the Internet community before our next meeting. We discussed a number of significant issues related to future automated key management mechanisms to support electronic mail privacy, and appreciated presentations by three guest speakers: Danny Cohen on a proposed privacy mode for TCP, Steve Kille on the developing state of CCITT directory service mechanisms, and Jim Hopkins on the state of the ISO security architecture addendum. The next meeting was tentatively scheduled for early 1987 on the West Coast. John Linn ROBUSTNESS AND SURVIVABILITY No report received. SCIENTIFIC COMPUTING No report received. SECURITY No report received. TACTICAL INTERNET No report received. TESTING AND EVALUATION No report received.