[IMR] IMR87-04.TXT Westine [Page 1]  ~ APRIL 1987 INTERNET MONTHLY REPORTS ------------------------ The purpose of these reports is to communicate to the Internet Research community the accomplishments, milestones reached, or problems discovered by the participating organizations. This report is for research use only, and is not for public distribution. Each organization 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 Ann Westine (Westine@ISI.EDU) or Karen Roubicek (Roubicek@SH.CS.NET). BBN LABORATORIES AND BBN COMMUNICATIONS CORPORATION --------------------------------------------------- SATNET The SATNET has continued to provide stable service throughout the month. We are continuing to experience some problems with channel 1 hardware. Tanum is experiencing problems with its modem and is now only on channel 0. We are still waiting for the spare modems from Linkabit to pass testing. Channel 0 has remained healthy so problems with channel 1 have not seriously affected the overall performance of the SATNET. Connectivity tests performed by Steve Casner reported an availability of 99.8% out of 99.9% for the SIMPs during the month of April. Westine [Page 1]  Internet Monthly Report April 1987 WIDEBAND NETWORK The development of both the dynamic group addressing and shared stream allocation features for the BSAT has been completed. These features were installed in the Wideband sites early in the month via the distribution of BSAT software Release 4.0. Remote monitoring of the BSAT's stream allocation data base was also initiated with Release 4.0. Assembly of the first run of Butterfly Satellite Modem Interface (BSMI) boards was completed during the month. Approximately one half of these boards were successfully tested and made available for installation in the BSATs. BSAT software Release 5.0, incorporating support for the BSMIs, was distributed to the field mid-month in preparation for the BSMI installations. At the month's end, BSMIs were being used for satellite channel access at the BBN and Lincoln Laboratory sites. Installation of the remaining BSMIs, including correction of a BSMI problem at CMU and Ft. Monmouth, will be performed next month. The Wideband Network was moved to a CONTEL ASC (aka American Satellite) transponder on the Westar-IV satellite on 24-25 April. The move was performed for contractual reasons only, CONTEL ASC having replaced Western Union as the provider of the Wideband Network's earth stations and space segment. The transition went smoothly, with no significant problems encountered. GATEWAYS The Butterfly and LSI-11 gateways continue to provide stable service. There are currently about 220 active networks in the Internet. We are working to add IP reassembly to the Butterfly and LSI-11 gateways to be able to handle fragmented EGP updates. This has already happened on Satnet and the updates will soon be too large for the ARPANET/Milnet. We have isolated the Ethernet problem that was showing up in the some of the Wideband Butterfly Gateways (ISI-WB, MIT-WB, SRI-WB) to a cable which was out of spec (too long). We are testing a shelded version of the cable which so far seems to work fine. Bob Hinden (Hinden@BBN.COM) Westine [Page 2]  Internet Monthly Report April 1987 ISI --- Internet Concepts Project Paul Mockapetris made draft revisions of the Domain RFCs 882 and 883 for the Internet Engineering Task Force meeting in Boston at BBN April 21-24. Paul worked with Doug Kingston of BRL in identifying some operational problems in the existing system and worked on the design of the Protocol Accelerator. One RFC was published: RFC 999: Westine, A.W., and J. Postel, "RFC 900-999 Request for Comments Summary". Paul Mockapetris (Mockapetris@ISI.EDU) and Ann Westine (Westine@ISI.EDU) Multimedia Conferencing Project and SuperComputer Workstation Communication MMC and SWC activities this month were joined in an interesting way: The multimedia conferencing facility was tested by actual use for discussions of both MMC and SWC topics. First there was a teleconference for collaboration between BBN and ISI on refinements to the new ST protocol document. We found that collaborative editting of the document in this mode worked well and have decided that future work on the document will be done with Diamond and MMCONF. The second teleconference was among MIT and ISI folks to review the results of recent experiments with the NETBLT protocol over the Wideband Net. MMCONF was used to examine relevant points from several typescripts and messages at both sites simultaneously. Alan Katz did further work on an X-Windows environment on the Sun as similar to suntools as possible. Annette Deschon and Alan did local tests on traffic generated by XDemos on the local ethernet and found a pretty constant rate of 50 packets per second each way (in some sense, this should be the maximum traffic each S un running X-Windows should be able to generate). Alan also did tests with the same environment running on DARPA's Vax over the wideband net and found performance slow, but acceptable, as was expected. Alan began learning about NEWS (and postscript) as an alternative remote window procedure (which may in fact be better for the supercomputer/workstation application, though it is now not as popular). Alan attended the IAB Scientific Computing Task Force meeting at the Space Telescope Institute in Baltimore on Westine [Page 3]  Internet Monthly Report April 1987 April 23-24. Brian Hung has automated the process of sending multimedia messages containing bitmap data from the IBM-PC AT using the old Multimedia Message Content Protocol(MMCP). Brian is working on a new program that performs the same function as the current multimedia scanner input program but uses the Diamond document format instead. Brian has written a small program that formats a message containing only header information and have sent out some messages to Steve Casner. Steve was able to receive the messages with out any problems. Brian's next step is to add bitmap data to the multimedia message. Steve Casner, Alan Katz, and Brian Hung (Casner@ISI.EDU) NSFNET Project Annette DeSchon distributed to NSFNET gurus a proposal for the data to be collected in an NSFNET gateways database. She has collected much of this data for the current NSFNET topology. In addition, she developed programs to automatically generate summaries of the gateway data and templates for the collection of additional data. From this data she generated an updated network-level map of NSFNET and also a subnet-level map that shows the internal structures of SURANET, NYSERNET, and JVNC. These maps were distributed at the meeting of the Internet Engineering Task Force at BBN. Bob Braden chaired a meeting of the End-to-End Protocols Task Force at CMU, April 15-17, and attended the INENG meeting at BBN, April 21-24. The revised gateway requirements document, authored by Braden and Postel, was completed in draft form, and made available for public FTP retrieval. Braden discussed the document at the INENG meeting. Bob Braden (Braden@ISI.EDU) and Annette DeSchon (Deschon@ISI.EDU) MIT-LCS ------- No internet related progress to report. Liixia Zhang (Lixia@XX.LCS.MIT.EDU) Westine [Page 4]  Internet Monthly Report April 1987 NTA & NDRE ---------- No report received. ROCKWELL INTERNATIONAL ---------------------- Tactical Internet Multicast Protocols David Young attended the End-to-End Services Task Force meeting on 17 Apr 87. He met the evening before with Steve Deering to discuss the extensions to IP multicast proposed in his (David's) draft RFC. The discussion of David's RFC at the task force meeting itself was primarily on his proposal to pass a membership list as a call parameter to expedite the host group formation. Jim Stevens could not attend the meeting, but he sent a network message to the task force describing new distributed methods to do multicast group assignment within the confines of the 32-bit IP address field. The key to the first method is allocating 12 bits of the 32-bit address to a unique number - the "serial number" - belonging to the multicast agent that is doing the assignment. The second method provides for "captive" assignment - that is, assignment, in the absence of a multicast agent, by a local host, using its own address as a portion of the address that it assigns. John Jubin (Jubin@A.ISI.EDU) SRI --- No report received. UCL --- UCL has now extended the local swamp to 3 ethernets, as well as the three rings, and the Admiral internet. The Admiral network is being upgraded to use 256kbps between each gateway interface, rather than 64kbps per channel. We are also connecting our workstations to an X.25 switch which provides up to 64 * 64kbps channels, as a fallback network for the department. The Eternets are subnetted together via a Cisco gateway. We are investigating the use of MAC bridges with access control and loop resolution for further extension of the local system. Westine [Page 5]  Internet Monthly Report April 1987 A preliminary graphical tool for investigating TCP and SEP performance over various network paths has been completed. John Crowcroft (jon@CS.UCL.AC.UK) UNIVERSITY OF DELAWARE ---------------------- 1. Development continues on the Dissimilar Gateway Protocol (DGP). Dave Mills simulated candidate multiple-path routing algorithms using a test jig written in BASIC and the configuration data previously collected for the NSFNET Backbone and client nets. As evidence the incredibly rich profusion of communal connectivity, the algorithms found up to eighteen distinct paths between some nets. Work continues on modifications to the algorithms to incorporate hierarchical constraints as proposed in the design documents. A design review was held at Linkabit on 9 April. 2. Dave Mills attended a NAS meeting on survivable telephone networks at CONTEL in New York City. His attendance at the joint INENG/ANSII meeting in Boston was cancelled at the last minute due to a nasty case of flu. 3. Gary Delp continues to do battle with the bits on Memnet. The hardware is growing state machine by state machine. Two technical reports have been put together, one describing an analysis of the board under construction, and the other an overview of the Memnet project. The analysis is co-authored with Adarshpal Sethi, and Dave Farber. The overview is written mostly by Marshall Rose, an adjunct faculty member. These reports are available from the Department of Electrical Engineering, University of Delaware, Newark, DE, 19716. 4. Memnet is a research project modelling the local area network resource as globally shared memory rather than a packet transport system. The processors perform reads and writes on locally cached, but consistent memory. Gary indicates that use of this distributed resource costs only a 10X performance hit over the "equivalent" non-shared resource; this is seen as a *major* improvement over software driven/packet network alternatives. 5. "A Distributed-Protocol Authentication Scheme," authored by Dave Mills, appeared as RFC-1004. This scheme was originally intended for use with DGP, but can be used with many similar applications. Reply comments indicate active and continuing interest in this area. Westine [Page 6]  Internet Monthly Report April 1987 6. Testing of the new fuzzware version for NSFNET Backbone sites and other places has turned up an elusive bug that results in occasional system crash/restarts, but only in some configurations. In spite of exhaustive testing the bug has not yet been found, although substantial review and rework of possible causes has been completed. 7. The WWVB radio clock at Linkabit has returned to life. Due to resource problems, the TCP-based time service TCP/TIME is being withdrawn from the fuzzballs. The UDP-based time services UDP/TIME and UDP/NTP will continue indefinately. Users have been warned on many occasions that this would happen. Dave Mills (Mills@UDEL.EDU) NSF NETWORKING -------------- UCAR/BBN LABS NSF NETWORK SERVICE CENTER (NNSC) We are working on setting up an on-line calendar of meetings of Internet-related groups such as task forces, working groups, etc., and are collecting information from chairpersons of such groups. Craig Partridge attended the End-to-end Task Force meeting and presented the state of work on the High-Level Entity Monitoring System (HEMS) to the Internet Engineering Task Force meeting. Drafts of some of the RFCs explaining the system are now being circulated for comment. Karen Roubicek attended the METANIC Meeting in Washington to discuss cooperation among the growing number of information centers. By Karen Roubicek (roubicek@nnsc.nsf.net) NSFNET BACKBONE SITES CORNELL UNIVERSITY THEORY CENTER Backbone Operations 21,225,000 packets were delivered by the backbone in March, up 26% from February (which was up 35% from January). The backbone fuzzballs are feeling the stress; besides the usual number of power problems and line outages one usually experiences, a number of software problems were uncovered and most of them fixed. The "dropped" statistics (counts of valid but undeliverable packets) have been discovered to be less than useful on some of the fuzzballs, although not all. It seems that every time certain Westine [Page 7]  Internet Monthly Report April 1987 fuzzballs emit a "Hello" packet to an attached ethernet the dropped count is incremented. We still don't know why. Coordination and Interoperability Issues Gatedaemon current events: Hold-down periods in all protocols are now rationalized; even more sanity-checking to avoid routing loops; interfaces are handled more effectively, including better support of point-to-point links. Also all copyrights except the original ones (Berkeley & Kirton) have been removed. UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN We have converted our campus to a class B backbone. It is our intent to control access to the Internet by only allowing that net Internet access. We have a connection to our the Chicago campus using P4200 at 56kb. They are in the process of acquiring a connected address to use at that campus. Another P4200 daisychained off that is 192.5.200 (Argonne National Labs) - who will be reachable via NSFNET as soon as we can do some gated'ing. TCP/IP and FTP have been running on CTSS for the past month crashing our Cray fewer and fewer times. FTP is currently open to all local users and is working reliably now. The service is unavailable from off campus until we convert the hyperchannel to a subnet of the connected net. No code to do - we just wanted to not introduce a lot of noise for initial trials. TELNET server was written to spec and requires local echo. Amazing how many clients don't support that. Lest people doing supercomputing think this junk at Illinois for CTSS is the only place they will have trouble, our implementation is almost exactly the same as UNICOS. Great minds think alike and having a Cray echo characters across a hyperchannel is not a pleasant thought. We have dropped off of USAN and are converting our Vita-Link dish to a hub. This conversion and the installation of a dish at Indiana University should occur about 15 May. By Ed Krol (krol%uiucvmd.bitnet@WISCVM.WISC.EDU) JOHN VON NEUMANN SUPERCOMPUTING CENTER No report received. NATIONAL CENTER FOR ATMOSPHERIC RESEARCH AND UNIVERSITY SATELLITE NETWORK PROJECT The Splitting of the UCARNET/USAN (128.117) into two nets (128.117 and 128.116) is now complete and has been stabilized. The 128.116 net will increase its functionality as a pass-thru network with the addition of a PROTEON P4200 IP router to NAS at NASA Ames (128.161) Westine [Page 8]  Internet Monthly Report April 1987 for the PLDS project of the NASA Science Internet (NSI). Another IP router box is anticipated between 128.116 and SESQUINET (Texas). Imminent is the arrival of a SUN 3/280 server that will gateway 128.116 to the NCAR HYPERchannel network and to an ARPANET PSN at the University of Colorado (56 Kbps). The NCAR HYPERchannel network uses internal protocols. Connectivity is currently available directly via 128.116 to the following nets: 128.117, 192.12.207, 128.121, 128.174, 192.5.146, 128.84, 128.8, 35, 128.193, 128.104, 192.12.220, 192.12.221, 10, 128.128, 18, 128.138. By Don Morris (morris@scdsw1.ucar.edu) PITTSBURGH SUPERCOMPUTING CENTER NSFNET The PSC Fuzzball and PSC-Gateway were relocated, and the lines to JVNC and UIUC reterminated on April 6. The line from PSC-Gateway to PSN 14 was extended at the same time. With the exception of the PSN, all of our long-distance communications lines now terminate in the same building as PSC offices, making it easier to keep an eye on things. There was a site survey on April 9 for a proposed new PSN, part of a larger plan to reduce ARPANET congestion. Installation is estimated to occur in August. The PSC Fuzzball handled 16.9 million packets in April. There are problems (first noted on April 17th) with performance and stability, possibly UDP related. We have a netwatch session running to try and see a problem. We will be evaluating a Hewlett Packard Ethernet Protocol analyser for general use, and will see if it can help here. PSC Fuzzball was rebooted on April 23rd and 28th at the request of the NISC, to fix links confused by these errors. PSCNET In view of the improved service to B.PSC.EDU via a dedicated path to PSC-Gateway, we have also installed a second Ethernet interface in A.PSC.EDU, which is now being announced as 128.182.65.1. Performance has been much improved, though there are still some problems with the VMS IP software. While we are waiting for bug fixes, we have adjusted parameters to minimize the impact. As a result of bringing PSC-Gateway and the Fuzzball to our building, we have been able to eliminate one of the gateways on the direct IP path. There are currently only 2 P4200s on this link, one at PSC offices and one at the Westinghouse machine room near the XMP. The gateway at PSC is on the stub Ethernet with the Fuzzball Westine [Page 9]  Internet Monthly Report April 1987 and PSC-GW, and the other is on the private Ethernet whose only occupants are A.PSC.EDU and B.PSC.EDU. The link between the two is running at 256kbps. ALCOA came on line this month, with a 56kbps connection to the P4200 at Westinghouse. We are announcing 192.31.3 to the Internet, their domain is ALCOA.COM. Lines have been ordered for the first link in the T1 Academic Affiliates net, sometimes called PSC-AA-NET. The first link is being installed to the University of Maryland in College Park. By Eugene Hastings, Mgr. Communications, PSC (hastings@morgul.PSC.EDU) SDSCNET SAN DIEGO SUPERCOMPUTER CENTER UC Santa Barbara, our first consortium member linked to the center over a tcp/ip-Proteon p4200 path, became operational during the month. While it took awhile to get the line "in shape" (it is a 56k DDS line with a cross-campus broadband link at EACH end), it is now functioning very smoothly. We expect to install Proteon's DECNET forwarding software about 1 June. While this will be of some use to UCSB, it will primarily serve one of our initial consortium members, the Salk Inst. plus an industrial participant. The site inspection for our ARPANET PSN was conducted during April by DCA/BBN/CRC. The PSN is still scheduled for an Aug. 1987 installation. During April we installed a new version of the SRI Multinet tcp/ip software. It has several improvements, one of the main ones being the elimination of the "tiny-gram" problem; the previous versions proclivity for sending ftp'd files a byte or two per packet. We are also in the midst of installing the LBL mail system SoftwareTools. When we complete the job in the next few weeks, mail from all of our networks will be gatewayed. These include, in addition to NSFNET, BITNET, SPAN, SDSCNET, MFENET and HEPNET. Corinne Carroll (ccarroll@SH.CS.NET) NSFNET REGIONAL AFFILIATED AND CONSORTIUM NETWORKS BARRNET BARRNET has been completing hardware installation during the past month. Much of this effort has involved a detailed learning process of T1 communications technology and in particular procuring and testing CSU's which maximize throughput of a random bit stream Westine [Page 10]  Internet Monthly Report April 1987 such as that produced by the Proteon p4200 gateways. At present 5 of our 6 initial nodes have had all hardware installed and tested (only NASA/Ames remains). An important concern that needs to be resolved before BARRNET is made fully operational is the control of routing information propagating out of each campus ("autonomous system") to other BARRNET gateways and associated campuses. We don't want bogus information to "damage" other parts of our networks and we don't want certain legitimate information propagated because of administrative or technical restrictions. We have outlined a proposed solution which involves a minimal software change by Proteon and anticipate that most of the nodes comprising BARRNET can be made fully operational by the end of May. No firm date is yet available for BARRNET's connection to NSFNET, although we plan to make this connection through NASA/Ames to the San Diego Supercomputer Center (SDSC). By Tom Ferrin (tef@cgl.ucsf.edu) MERIT Staff training for use of the CMU enhanced asynchronous version of the MIT PC/IP package has started and we now have generally many more users of the asynchronous IP support on Merit terminal lines. Preparation for general user distribution of floppy disks for the async IP/TCP support as well as the Ethernet version has continued as well. The Ethernet support is still beta tested by a couple of sites, which are using the support fairly heavily, including for the exchange of multimedia mail documents. Since the IP switching function in Merit is working to a large extent (more still needs to happen for congestion control) work has started on a TCP within the Merit network for terminal as well as host support. To this extent the interface to our MTS mainframes is also now being worked on, but further design discussions still need to happen. Hans-Werner Braun attended the Internet Engineering Task Force meeting in Boston and also visited Proteon together with Milo Medin (NASA) and Mike Petry/Louie Mamakos (UMaryland) on the preceeding day to discuss several aspects of networking issues with regard to Proteon efforts. Dave Katz attended a meeting of the ANSI X3S3.3 (Network and Transport Layer) group, where he is a voting member. By Hans-Werner (hwb@mcr.umich.edu) Westine [Page 11]  Internet Monthly Report April 1987 MIDNET Our gateway/routers have arrived; we are still waiting for delivery of the DSU's/CSU's. We have scheduled a training seminar for mid June to cover the operation of the gateway/routers. Telephone lines are still scheduled for June 1. We are in the process of establishing a MIDNET NIC. By Doug Gale (doug%unlcdc3.bitnet@WISCVM.WISC.EDU) NORTHWESTNET No news to report from last month. We are still waiting for the official NSF approval. By Hellmut Golde (Golde%UWACDC.Bitnet@WISCVM.WISC.EDU) NYSERNET As of 1 May 1987, NYSERNET has the following topology with 56kbit links and Proteon Gateways. Reachability to NSFNET, ARPANET, MILNET, etc is available to each site. April additions include SUNY Albany, Clarkson University, and NYNEX Science and Technology Center. Clarkson | | Rochester--------Cornell---------RPI--------SUNY/Albany | | | | | | NYTEL----Columbia------NYU | | | | | NYNEX | BNL (Brookhaven National Laboratories) The current implementation schedule has all nysernet academic members available via 56kbit service by the 3rd week of May. T1 cutover for NYU to Columbia and Columbia to Cornell is schedule for June/July. Creation of an experimental root domain server is in progress. By Marty Schoffstall (schoff@nic.nyser.net) SESQUINET We are in the process of ordering gateways (from Cisco) and phone lines (from SouthWest Bell), and plan to be up by June 15th in our Westine [Page 12]  Internet Monthly Report April 1987 initial configuration of: Baylor College of Medicine Houston Area Research Center Rice University Texas A&M University Texas Southern University University of Houston at University Park We are also working with the NSFNET people to build a direct connection from the Rice node to the NSFNET backbone site at NCAR. We are, as ever, open to contacting new potential members, but our focus during May will be in getting our initial configuration up. Questions regarding growth or operations can be sent to Guy Almes (almes@rice.edu) or to Farrell Gerbode (farrell@icsa.rice.edu). By Guy Almes (almes@RICE.EDU) SURANET The following nets are being EGP advertised to the core on SURANET's behalf. 128.61. Georgia Tech 128.109. TUCC 128.150. NSF 128.154. NASA Goddard 128.163. U of Kentucky 128.164. George Washington Univ 128.167. SURAnet 128.169. U of Tennessee 128.173. Virginia Tech 192.5.57. Univ of Delaware (udel-cc) 192.5.219. Clemson 192.16.177. Univ of Alabama The phone line from Georgia Tech (Atlanta) to the University of Georgia (Athens) has been installed. A Proteon router has been shipped from Maryland to the University of Georgia and is awaiting installation. NASA Goddard (128.154.) was connected to SURANET this month. Discussions are going on with other Federal research laboratories about connecting to SURANET. Reliability of the Proteon p4200 boxes, using Version 7.2 of the Proteon software and the resistor fix, continues to be good. Using software obtained from Charles Catlett at the University of Illinois we plan to initiate 24 hour monitoring of SURANET node Westine [Page 13]  Internet Monthly Report April 1987 reachability in the near future. We are contacting telecommunication suppliers in preparation for connecting the ten SURANET Phase II campus sites to the network. Video tapes of the Feb. 13-14 SURANET Technical Meeting are available by sending a P.O., or check, for $91.00 to SURA Inc. c/o Jack Hahn, Computer Science Center, University of Maryland, College Park, MD., 20742. SURANET will be holding a Network User Services forum primarily for SURA members but open to other interested NSFNET members. For information contact Jack Hahn (@umd2.umd.edu:hahn@umdc.bitnet) or Henry Schaffer (tsches@tucc). SURANET is testing a conferencing server concept being operated by Dr. Jeff Case at the University of Tennessee using Vaxnotes over SURANET. By Jack Hahn (hahn%umdc.bitnet@WISCVM.WISC.EDU) WESTNET A letter was transmitted to NSF on April 2, 1987, entailing a detailed response to the questions raised by NSF regarding the proposal submitted in September of 1986. This also included a revised budget. Issues still remaining include: (a) 56 kbps access by Utah State University and the University of Wyoming. It appears that the University of Wyoming will have 56 kbps access through a subchannel of a microwave link which will be installed during the forthcoming Colorado fiscal year (July 1, 1987 to June 30, 1988). The circuit to Utah State is still questionable, and expensive (over $3,700 per month). (b) Interaction with the backbone. It appears that the backbone node (in our case NCAR) is to be responsible for the circuit(s) to the regional networks. I perceive this as a potential problem, as the installation of this circuit must be done in a timely manner, else the region will be inter connected and waiting for external connectivity. I recommend that the regions be given responsibility for these circuits. However, I do agree that the backbone node should be responsible for all hardware beyond the transceiver cable at the backbone node end. (c) We as yet have no funding. This is beginning to evoke severe reaction at the regional sites, as many have invested heavily in campus LAN's and feel as if they have done their part Westine [Page 14]  Internet Monthly Report April 1987 regarding purchase of hardware, and NSF should 'deliver' its part. Furthermore, many researchers throughout the region presently have a claimant need to access Phase II Centers NOW... Progress during the last month, in addition to the response to NSF, includes: (a) a thorough circuit definition and cost analysis, (b) completion of Colorado Supernet, (c) connection through NCAR to NSFnet by Colorado Supernet (d) the formulations of: an initial marketing plan, an advertising brochure, and a one page fact sheet. By Pat Burns (pburns%CSUGREEN.BITNET@wiscvm.wisc.edu) Westine [Page 15]  Internet Monthly Report April 1987 TASK FORCE REPORTS ------------------ APPLICATIONS -- USER INTERFACE Current members are myself and Richard Bolt (MIT Media Lab), Steve Casner (ISI), Terry Crowley (BBN), Jose Garcia Luna Aceves (SRI), Sunil Sarin (CCA), Joseph Sventek (ANSA, Cambridge, UK), and Dave Taylor (HP Labs). Keith Lantz (Lantz@SCORE.STANFORD.EDU) AUTONOMOUS NETWORKS Informal discussion notes from the March meeting available from estrin@usc-oberon.usc.edu. Distributed copies of Dave Mills' DGP and the ANSI Draft Network Routing Architecture document for online discussion. Deborah Estrin (Estrin@USC-CSE.USC.EDU) END-TO-END SERVICES The END2END task force met for two days at CMU during April. Three new members were present -- Van Jacobson, Bill Nowicki, and Craig Partridge. Dave Young of the SURAN project at Rockwell also attended for one day, to talk about his proposed extensions to the Internet multicasting facility to support specific SURAN requirements. Van gave a presentation of his work on TCP/gateway performance issues; Joel Emer talked about the MIT Common System project; and Craig discussed the transport protocol issues in gateway monitoring. There is now a multicast agent implementation for 4.3BSD; it will be available as soon as some BBN licensing issues are straightened out. Deering has written an RWHO daemon that use Internet multicasting instead of broadcasting (!). The Internet multicasting code is nearly working in MACH at CMU. Cheriton reported progress on a BSD kernel implementation of his VMTP protocol; it implements many features of VMTP, including: logical name resolution (ESP), authentication, multicasting, and forwarding. One potential VMTP application which Dave is considering is encryption key distribution. The task force decided that VMTP would also be an appropriate Westine [Page 16]  Internet Monthly Report April 1987 transport protocol for gateway monitoring. To this end, Cheriton plans to develop a VMTP subset simple enough for inclusion in gateways, and try it out monitoring gateways in the Stanford environment. Van Jacobson made a number of interesting points, based on his analysis of the coupling between TCP retransmission and the gateway system: (1) The variances are as important as the means. (2) There is an inevitable tendency (predicted by Catastrophe Theory) for the system to organize itself to oscillate between empty queues and full (or overflowing) queues. The most benign result is that packets always get clumped. Van uses physical methaphors to describe this, like "a growing packet crystal" and "cooling the packets". (3) His experiments and simulations show that this tendency can be controlled with either queue randomization or fair queueing in the gateways. (4) RTT is the fundamental transfer function for the gateway system. (5) Measures which have been incorporated into host TCP's for CPU efficiency, like sending large bursts and economizing on ACK's, may be a bad idea from the viewpoint of throughput. (6) His "slow start" algorithm greatly improves throughput. There was a vigorous discussion on point (5); several participants felt that clumping, both in the hosts and in the network, should be a good thing, as it improves efficiency. For example, clumping tends to make better use of ARPANET connection blocks (and the VC setup in the ARPANET in turn probably helps cause clumping). It was unclear how applicable Van's current results are to rate-controlled protocols like NETBLT. Van admitted that while he knows now how to make TCP self-tuning, he does not know how to make NETBLT self-tuning. It may require more explicit performance data from the gateways to the hosts. The idea was advanced that the gateway system ought include an explicit internal organization to manage its overall buffering as a resource. Bob Braden (Braden@ISI.EDU) Westine [Page 17]  Internet Monthly Report April 1987 INTERNET ARCHITECTURE The INARC list was quiet this month. Dave Mills (Mills@UDEL.EDU) INTERNET ENGINEERING 1) A special meeting was held in early April to discuss Internet technical coordination between various agencies (including DARPA, DCA, NSF, NASA and OSD). The invited attendees were: o Mike Corrigan, OSD o Dennis Perry, DARPA o Steve Wolff, NSF o Phill Gross, MITRE o Mike St.Johns, DDN PMO o Hans-Werner Braun, U. Michigan o Milo Medin, NASA In order to deal with a potentially very large combined technical group, initial discussions centered around a multiple working group format in which there could be a small central coordinating committee and a varying number of ad hoc and standing working groups. It is envisioned that the working groups will have a life of their own, with meetings possibly scheduled outside the normal IETF. Future IETF meetings would be composed of detailed working group status reports, other technical presentations and an opportunity for the working groups to meet. As an example, a three day meeting might dedicate the first day and a half to working groups, the remainder dedicated to a plenary session of technical presentations and status reports. It is expected that the Working Group Chairs will report progress monthly (to be included in this report and/or a separate Task Force report) as well as at the full IETF meetings. 2) The Internet Engineering Task Force met in Boston on April 22-24. The meeting was hosted by Bob Hinden of BBN. Bob and BBN did an excellent job of providing facilities for what turned out to be a rather large gathering. This was a very large meeting for two reasons: this was the first meeting since the IETF reorganization (described above) and it also included a day and a half joint session with the ANSI Network Layer Group X3S3.3. An attendence list has been distributed to the IETF mailing list. Westine [Page 18]  Internet Monthly Report April 1987 There were 47 attendees who listed IETF affiliation, 31 who listed X3S3.3 affiliation (including several who listed dual affiliation) and 10 from agencies (and/or BBN observers) who listed no affiliation. This gives a total of over 85 attendees. 3) The final agenda for the April 22-24 meeting is below. Thanks go to Lyman Chapin who helped coordinate the joint session of the meeting. Wednesday, April 22 ------------------- Morning - Welcome, Task Force Reorganization Gross (MITRE) - Enhanced AHIP StJohns (DDN) - BBN Report Hinden/Gardner (BBN) - Progress Report on - Congestion Control Simulation Stine (MITRE) - Arpanet Performance Measurement Gross (MITRE) - TCP Performance Enhancement Jacobson (LBL) Afternoon - Gateway Monitoring Partridge (BBN) - Management Architecture LaBarre (MITRE) - Internet Problem Descriptions Groups Thursday, April 23 ------------------ Joint X3S3.3/IETF Meeting on Gateways and Routing (concluded Friday Morning) Morning - Welcome Chapin/Gross - IETF status/overview Gross (MITRE) - FCCSET report Gross (MITRE) - ANSI/ISO status/overview Chapin (Data General) - ANSI routing architecture Tsuchiya (MITRE) - NSF gateway requirements Braden (ISI) - Routing Directions at SRI Su and Garcia (SRI) Afternoon - Burroughs Integrated Adaptive Routing Piscitello (Unisys) - DECNET Phase V Routing Oran (DEC) - Discussion & questions Westine [Page 19]  Internet Monthly Report April 1987 Friday, April 24 ---------------- Joint X3S3.3/IETF Meeting on Gateways and Routing (Con't) Morning - SPF Routing in the Butterfly Gateways Mallory (BBN) - Other Advanced Routing Work at BBN Gardner (BBN) - Congestion Avoidance Jain, et al (DEC) - Adjourn Joint Session Afternoon - Parallel IETF Working Groups 4) There were two breakout sessions during the IETF meeting. On Wednesday afternoon, separate groups worked on Internet problem descriptions and, on Friday afternoon, working groups were organized. Internet Problem Descriptions At the first IETF meeting in January 1986, a list of Internet problems was developed covering short, intermediate and long range issues. In response to IAB guidance, an attempt was made to develop such a list in a more rigorous fashion at the February 1987 meeting. The IETF membership was divided into groups with the goal of compiling problem descriptions in particular areas. Problems were listed in the following format: o Problem Category o Problem Description: o Severity: (low, medium, high) o Time Frame: (time until problem becomes critical) o Owner: (Responsible Agency or group) o Plan/Options: The resulting Internet Problem Descriptions (designated "IETF Form-1") contained a mixture of intermediate range protocol issues and very short range O&M issues. The resulting problem list was not exhaustive and further work was done on Wednesday afternoon, April 22. These new forms will be edited and made available in the IETF archive directory at SRI-NIC. Task Force Working Groups Working Groups should have a narrow focus. Ideally, they should be formed to address a single issue and then disolve when the work has been accomplished. With this philosophy in mind, working group topics were identified and appropriate chairs were chosen during the course of the meeting. On Friday afternoon, the following groups met: Westine [Page 20]  Internet Monthly Report April 1987 - EGP2 RFC (Petry, UMd) - Name Domain Planning (Kingston, BRL) - Gateway Monitoring (Partridge, BBN) - Performance and Congestion Control (Stine, MITRE) - NSF Routing (Hedrick, Rutgers) - Misc. MilSup Issues (StJohns, DDN PMO) Reports have already been received from most of the Working Groups. These reports will be compiled and sent to the task force mailing list. Since this was a preliminary meeting for these groups, there may be further refinement in their specific charters. 5) Mike StJohns has arranged for an archive directory at the NIC for task force papers. The directory is on the host SRI-NIC. All files are available for anonymous FTP. 6) Using the attendence sheets from the last meeting, the Task Force mailing list has been checked, corrected and updated. The main list, ineng-tf, has 97 entries. The interest list has an additional 21 more. With a list this large, it turns out to be very convenient to compile and edit it locally. Therefore, I have created three new lists at MITRE: - - ineng-tf@gateway.mitre.org (aliased to ietf@gateway.mitre.org) - - ineng-interest@gateway.mitre.org (aliased to ietf-interest@gateway.mitre.org) - - ineng- request@gateway.mitre.org (aliased to ietf- request@gateway.mitre.org) I plan to 1) continue to maintain the lists locally and 2) send copies of the full list to ISI at regular intervals. Submitted by Phill Gross (Gross@GATEWAY.MITRE.ORG)) PRIVACY The Privacy Task Force held a successful meeting on 31 March-1 April at RIACS, NASA/Ames, Mountain View, CA. The attendees were David Balenson (NBS), Matt Bishop (RIACS, meeting host), Morrie Gasser (DEC), Steve Kent (BBNCC, task force chairman), John Linn (BBNCC, task force secretary), Dan Nessett (LLL), Mike Padlipsky (Mitre/Bedford), Rob Shirey (Mitre/Washington), and Steve Wilbur (UCL). We began the meeting with discussion of the PTF's released Westine [Page 21]  Internet Monthly Report April 1987 RFC989, also discussing potential implementation activities and considering some issues raised by the minutes from the September 1986 meeting. Several PTF members are considering prototype implementation efforts based on the RFC; continuing discussion has identified some changes and clarifications to be made in a subsequent revision. We explored key management architectures to support electronic mail privacy, emphasizing a certificate mechanism based on RSA public-key cryptography. Major issues included definition of user identification fields to be included in certificates and consideration of possible server facilities to register users, generate certificates, and make certificates available to mail senders. The next meeting was tentatively scheduled for 28-29 July at NBS. John Linn (Linn@CCY.BBN.COM) INTERNET MANAGEMENT The first Internet Management Task Force meeting will be in May. Vint Cerf (Cerf@A.ISI.EDU) ROBUSTNESS AND SURVIVABILITY No report received. SCIENTIFIC COMPUTING We met at the Space Telescope Science Institute at Johns Hopkins University (hosted by Peter Shames) on April 23-24. Discussions were held in four areas: equations and multimedia communications, the Interagency Research Internet, the need for very high bandwidth communications, and network related user services (such as white pages and database directories). A demonstration was given of the Diamond multi-media mail system by Terry Crowley of BBN. Vint Cerf (Chairman of the IRI Engineering Task Force) joined us for the discussion on the Interagency Research Internet. Westine [Page 22]  Internet Monthly Report April 1987 We plan to generate three short white papers (perhaps to be published as RFCs). One will be on the different methods for equation input and interchange, and will be a follow-up to the RFC published by Alan Katz. The second will be a summary of the requirements for high bandwidth networking in the scientific community. The third will be on some of the network related user services needed in the scientific community and approaches to providing those services. We had a short discussion on the charter of the task force, and concluded that the title "Scientific Computing" was too narrow, as many of the future networking requirements for the scientific community will come from uses other than just computing, such as remote control and monitoring of experiments. We therefore have changed the name of the task force to "Task Force on Scientific Requirements." Barry Leiner (Leiner@ICARUS.RIACS.EDU) SECURITY No report received. TACTICAL INTERNET No report received. TESTING AND EVALUATION No report received. Westine [Page 23]