[IMR] IMR86-11.TXT Westine [Page 1]  ~ NOVEMBER 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 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 --------------------------------------------------- VAX UNIX NETWORKING In the month of November, the kernel-level implementation of the Internet Multicast Agent for 4.3 BSD Unix was almost completed. Karen Lam spent 10-21 November at Stanford University, testing and Westine [Page 1]  Internet Monthly Report NOVEMBER 1986 debugging the 4.3 BSD Multicast Agent against the Stanford software. Testing proved highly successful. The 4.3 implementa- tions are now able to run independent of the Stanford software. Stanford University shipped a Sun workstation to BBN in order for inter-agent testing to continue at BBN. Eric Cooper, of Carnegie- Mellon University, has volunteered to act as a beta test site for the 4.3 IGMP and Multicast Agent software. WIDEBAND NETWORK Significant time was spent this month debugging and improving the performance of the Wideband satellite channel equipment. Using data obtained by automatically monitoring the time variations of Wideband sites' power levels as received at BBN, a decision was made to repoint the antennas at all of the sites. The repointing must be performed when the satellite is positioned at the midpoint of its periodic motion relative to the Wideband earth stations. The repointing operation does not interfere with normal network operations and should improve network performance. The BBN, ISI, and M/ACOM-LINKABIT sites have been repointed as of this report. BBN is assisting American Satellite's efforts to determine the source of problems that result from the use of a control/status connection between the Earth Station Interface (ESI) and the earth station equipment. The cable used for this connection, previously disconnected at all Wideband sites, has been reinstalled at the ISI site as part of this investigation. BBN provided support for NETBLT protocol experimentation between MIT and ISI via the Wideband Network. An unusually high rate of packet loss was observed during these experiments. Debugging efforts have isolated a problem in the Interface Controller-CODEC Unit of the ISI Earth Station Interface which is believed to be a major contributor to the packet lossage. Spare ESI hardware is being shipped to ISI to correct the problem. Work on the BSAT's stream scheduling/synchronization software con- tinued during the month. The distributed stream scheduling data- base has been reorganized to be more efficient and far simpler than it had been previously. Multisite testing of the software managing this new database has been performed on the satellite channel and in a simulated all-digital network. A BSAT release incorporating the new software will be made soon. SATNET In general, the SATNET remained stable, providing continuous ser- vice on channel 0. However, toward the end of the month, Fucino's channel 1 transmissions began to interfere with the channel 1 Westine [Page 2]  Internet Monthly Report NOVEMBER 1986 transmissions of the other sites and its satellite modem had to be looped off the air. In light of the increased problems, we plan to go ahead and try to adjust the Fucino satellite equipment in December, even though we do not yet have the spare Linkabit modems. There were 3 scheduled outages during which the SATNET measurement taskforce conducted some experiments. GATEWAYS The month of November has been a combination of good and bad. The good news was that since new software (rel. 3.5) was installed, most of the Butterfly Gateways have been stable. The CSS Gateway, which we had been previously having problems, was up solidly for the whole month with out any outages. Except for one restart, NTA has been up since the new software has been installed. The other good news was that we found a bug which causes the Wideband gate- ways to restart a lot. The fix was patched into ISI-WB, which was the worst, and that gateways is now stable. The bad news is that RSRE has been plagued by hardware problems. After running for almost a week with the new software it got very unstable. This appeared to be fixed by doing a PM and updating the microcode on one processor. After the PM, it was up for about a week without any problems. Then, on the day after Thanksgiving, it broke solidly. We are sending over replacement hardware and hope to have it up soon. UCL, when it hasn't been isolated by the RSRE problems, seems to be running fine. Bob Hinden ISI --- Internet Concepts Project Two RFCs were published: Reynolds. J.K., and J. Postel, "Assigned Numbers" RFC 990, USC/ISI, November 1986. Reynolds. J.K., and J. Postel, "Official ARPA-Internet Proto- cols " RFC 991, USC/ISI, November 1986. Westine [Page 3]  Internet Monthly Report NOVEMBER 1986 Multimedia Conferencing Project Brian Hung has written a program for the IBM-PC AT to compose Multimedia messages containing bitmap data. He plans to test the validity of Multimedia messages generated by this program by sending it to a Sun workstation running the BBN's Diamond Multimedia system. Currently Brian is making preparations for this test. Joyce Reynolds presented a BBN Diamond Multimedia Mail demo to Peter Evans, of TELECOM Australia. Brian Hung and Joyce Reynolds Supercomputer and Workstation Communication Project Annette Deschon attended a meeting at XEROX in Leesburg, VA, 10-11 November. Alan Katz attended a Scientific Computing Task Force meeting at Lincoln Labs in Livermore, CA, 12-14 November. Testing of the NETBLT protocol over the past several months has been impaired by packet losses within the Wideband Network on transmission from MIT to ISI. The problem has been iso- lated to the ESI codec at ISI (part of the WBNET node), and corrected by replacement of some defective boards. An invalu- able aid in this testing was Bob Braden's addition of bit error checking to the ICMP ping program to illustrate the nature of the problem. Further diagnosis was made with the Bit Error Rate tester recently built into the BSAT software, with BBN and ISI folks working together at the controls. Steve Casner MIT-LCS ------- We are continuing the test of NETBLT over the wideband satellite channel with ISI. An implementation bug was recently discovered. NETBLT uses multiple concurrent transmission buffers, each has its own timer. The bug caused the processing time of the timers going up with the square of the number of buffers being used, hence explained some mysterious packet losses observed in previous tests. New throughput measures (after the bug fix) will be taken soon. We may see some improvement from the previous throughput value of 390 Kbps. Lixia Zhang Westine [Page 4]  Internet Monthly Report NOVEMBER 1986 NTA & NDRE ---------- No report received. SRI --- No report received. UCL --- No report received. UDEL ---- 1. I continued to collaborate with Linkabit staff on the design of the Dissimilar Gateway Protocol (DGP) for Ford Aerospace and RADC. I reviewed the latest ANSI Draft Routing Architec- ture document and concluded this a good basis for the DGP architecture. I presented the framework for a possible DGP design at a meeting with the principals and began filling in the details for a white paper to circulate among the task forces. 2. I continued tinkering with the Network Time Protocol (NTP) fuzzcode and smote a couple of nasty bugs that keeled some machines when Rich Wales' innocent NTPgrams dropped in from UCLA. Hosts there and on FORDNET, UMICHNET, UMDNET, DCNET and here at UDEL now derive network time either from local radio clocks or from remote clocks via NTP as backup. Plans are in hand to install a radio clock for the NSFNET Backbone and to extend the NTP service to its client networks. 3. During the month two of the three primary radio clocks (FORD1 and UMD1) died and have not yet returned to life, so the erstwhile DCN1 clock remains the only ticker in town and ever- ybody else tocks NTP to it. For a short time even that watch lost its spring, so the scruffy WWV clock on my desk became the master timepiece for the universe. While despairing for the bum clocks, I am very pleased about the overall system robustness and the performance of the distributed-peer algo- rithms, which are providing local time via NTP accurate to a few milliseconds, even across the ARPANET. Westine [Page 5]  Internet Monthly Report NOVEMBER 1986 4. Dusted off the supporting software which led to RFC-889 on Internet delay experiments and flew some pings across various networks and gateways. I then summarized in a message to TCP/IP the results, which include several interesting scatter diagrams showing the delay-length characteristic for selected ARPANET, MILNET and NSFNET paths. The diagrams are in the form of standard bitmap files that can light up a Sun. I even lit a Trailblazer packet-ensemble modem, discovered why it doesn't work well with packets, and sent a TCP/IP note on that. 5. While pinging the NSFNET swamps, I found a possible problem in a dinky X.25 circuit between the PSC-GW gateway (a microVAX running Ultrix and ACC 5250 X.25 interface) and its IMP. The problem results in excessive delays and delay dispersions and is thought due to bugs in the interface firmware or driver software. 6. Our UDEL SURANET gateway has been installed, but not yet hump- ing packets. A new Class-B network number was obtained for the UDELaswamp and a new Class-C number for Linkabit, clearing the way to rehome the restless DCNET swamp creatures to Newark. The zoo will be gatewayed to ARPANET using a borrowed port on IMP 96. 7. NSFNET now has an official gateway PSC-GW to ARPANET, so dinky DCN-GATEWAY at Linkabit no longer is sloshing upwards of a megapacket per week. However, much of the slack is being taken up by its other clients, as well as furnishing a backup to PSC-GW. During the month the NSFNET grew several new client networks and triggered a real shocker when it was discovered that the fuzzball routing packets were just a few octets shy of buffer overflow (in fact, FORDNET did overflow and all its fuzzballs crashed). DEFCON 5 was averted only by an Emergency Action Message that resulted in larger buffers. Our BBN tren- chstalkers will certainly enjoy that one. Dave Mills TASK FORCE REPORTS ------------------ APPLICATIONS No report received. Westine [Page 6]  Internet Monthly Report NOVEMBER 1986 END-TO-END SERVICES No report received. INTERNET ARCHITECTURE 1. There has been no significant mail activity on the task force list; However, there has been a good deal of activity on other lists, in particular TCP/IP, NSF- ROUTING, and in ad-hoc exchanges. The issues being hotly debated have to do with scenarios where gateways are shared between two or more autonomous systems, where more than one IGP is run in a system and where systems include gateways and protocol implementations from multiple ven- dors. The NSFNET community, presently being singed by these rapidly engulfing brushfires, has the highest octane fuel for the fires. 2. There is a critical need for further development of the Internet architecture to handle the exploding growth of the community and to provide needed improvements in ser- vices, especially in the area of interoperable routing systems. The draft routing architecture being discussed in ANSI committees in fact seems quite close to conven- tional Internet wisdom (even if that wisdom isn't always followed). In the face of real and immediate needs, many ad-hocs (interoperable routing daemons, for example) are being nailed to the woodwork without systematic planning, analysis and testing. The charter of the task force is to aid and abet the research and development process leading to these goals, which involves sticking noses in development projects like SURAN, MSS and NSFNET, lifting technology wherever found and submitting white and off- color papers for discussion. This points up the urgent need for the task-force members to take a participatory role in the discussions on other lists and to send alerts to the task-force list when appropriate issues come up. To that end, the chair invites the members to submit items of general interest for posting to this monthly report. Dave Mills Westine [Page 7]  Internet Monthly Report NOVEMBER 1986 INTERNET ENGINEERING No report received. INTEROPERABILITY The Autonomous Systems Task Force is still undergoing reorgan- ization. A draft document on Research Issues in Autonomous Systems is available but is due to undergo more revision before general distribution. If you are interested, or poten- tially interested, in this task force, or in providing feed- back on the draft document, please contact Deborah Estrin (estrin@usc-cse.usc.edu). PRIVACY Limited Privacy Task Force activity took place in the month of November. Some work on revising the Electronic Mail Privacy Enhancement RFC continued. John Linn ROBUSTNESS AND SURVIVABILITY No report received. SCIENTIFIC COMPUTING The second meeting of the Task Force was held at Livermore Labs on 14-15 November. The meeting concentrated on discus- sions related to a paper to be presented at an upcoming workshop. The paper will discuss the near term and long term goals for the functions to be provided by a network supporting scientific research. Topics were divided into four categories: connectivity requirements, network functions, net- work specifications, and user services. Discussions took place on each of these areas, and volunteers were solicited to write each of the sections. A draft of the paper is to be completed by the middle of December. Barry M. Leiner Westine [Page 8]  Internet Monthly Report NOVEMBER 1986 SECURITY No report received. TACTICAL INTERNET No report received. TESTING AND EVALUATION No report received. Westine [Page 9]