[IMR] IMR86-04.TXT APRIL 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@USC-ISIB.ARPA. Reports are requested from BBN, LINKABIT, ISI, LL, MIT-LCS, NTA, SRI, UCL, and UCLA. Other groups are invited to report newsworthy events or issues. BBN LABORATORIES AND BBN COMMUNICATIONS CORPORATION .Grab=5; --------------------------------------------------- WIDEBAND NETWORK On 1 April a demonstration of simultaneous packet video and packet voice across the Wideband Network was held. This demonstration supported a cross-country video conference to discuss multimedia mail. During the previous week, packet video equipment was moved from Lincoln Labs to BBN. Also during that week, software work on the BSAT stream facility (low delay TDMA service) was completed and used for the first time in the Wideband Net. The video conference successfully used the stream facility for the demonstration on 1 April. Also during April, the CMU and SRI sites were returned to working order. There are now eight operational Wideband Net sites. DIAMOND MULTIMEDIA COMMUNICATIONS SYSTEM We have started to schedule a regular weekly time slot (Thursdays, 1pm-3pm ET) for holding multimedia conferences between BBN and ISI using the BBN Diamond Multimedia Conferencing System, the ISI Packet Video System system and the BBN/ISI Packet Voice System system. All three systems use the Wideband Network for data transfer. Initially we will be using the conferencing facility ourselves, however after a while, we will be interested is letting others evaluate the system and give us their reactions and suggestions. People who are interested in testing this facility should contact Harry Forsdick (Forsdick@BBNA.ARPA) or Steve Casner (Casner@USC-ISIB.ARPA). GATEWAYS Release 3 of the Butterfly Internet Gateway software was installed in all of the Butterfly Gateways. This release of software is the first to contain dynamic routing (SPF and EGP). Also included in the release is Satnet HDH support, improved monitoring and control, improved Arpanet support, interface up/down, and bug fixes. We installed the first two Butterfly Gateways on Satnet, one in the U.S. at CSS and one in Europe at CNUCE (in Pisa, Italy). The gateway for NTA will be installed in the middle of May. The gateways for the U.K. will be installed shortly there after. There are now 10 Butterfly Internet Gateways installed. They are: BBN-UCC, BBN-WB, CMU-WB, CNUCE, CSS, IPTO, ISI, MIT, MIT-WB, and SRI-WB. SATNET During April, SATNET service continued to be stable. The channel 1 module at Fucino was fixed. Also, joint efforts by Stan Rothschild (COMSAT) and Kim Kaiser further improved the channel situation. They took the first step toward correcting the PSP terminal adjustments by tuning the frequencies used by the hardware. Tanum continued to lack channel 1 service, but retained internet connectivity and service via channel 0. A major milestone was reached when CNUCE in Pisa Italy was gateway'd to the SATNET. The SIMP at Roaring Creek was upgraded with an MSYNC board and together with the SIMP at Fucino received the HDH/HDLC software needed to support the new Butterfly gateways. The new gateways were then installed at CSS and CNUCE. Bob Hinden ISI .Grab=8; --- Internet Concepts Project Greg Finn finished two reports on a new routing mechanism for packet switched networks which has several advantages over currently used algorithms. In particular the ability to scale upward indefinitely and a relatively transparent treatment for mobile hosts, which are treated no differently than non-mobile hosts as far as intermediate routing nodes are concerned. Annette DeSchon converted the ISI Internet protocol software, which runs on the XEROX Dandelion workstation, to run under XDE version 4.0. The problems that prevented the Dandelions from receiving odd-byte-length packets and packets greater than 576 bytes in length have been solved. These problems were solved by building new bootfiles which incorporated several XEROX bug fixes. In addition, she developed an enhanced version of the XEROX Ethernet SpyTool, which is being used for performance analysis of TCP based FTP over the wideband network. This SpyTool will also be used to analyze the performance of new protocols such as the NETBLT protocol, which is currently being developed at MIT. Jon Postel edited two RFCs that were published this month. RFC 982 -- ANSI, "Guidelines for the Specification of the Structure of the Domain Specific Part (DSP) of the ISO Standard NSAP Address". RFC 983 -- Cass, D.E,. Rose, M.T., "ISO Transport Services on Top of the TCP". Ann Westine Multimedia Conferencing Project During the successful multimedia teleconference held on April 1 and reported last month, the Wideband Net was operating with only the ISI and BBN sites up and running an unreleased version of the BSAT software. Now the equivalent version has been released to all sites so it is once again possible to run packet video using stream transmission. With the packet video rate and stream parameters tuned to allow operation with all sites present on the network, we are now ready for routine operation of packet video and multimedia conferencing. We have begun a series of planned weekly teleconferences to test the system's effectiveness. Small groups in the vicinity of ISI and BBN are invited to try telecommunicating with each other by signing up for one of the weekly slots. Contact Casner@USC-ISIB.ARPA or Forsdick@BBNA.ARPA. Brian Hung is working on expanding the existing bitmap file produced by the IBM-PC AT linearly by one half so that the resulting bitmap file will fill an entire 11" X 8 1/2" page when output on the Imagen. This is due to the fact that the Microtek MS-200 scanner has a 200 dot per inch resolution whereas the Imagen printer has a 300 dot per inch resolution. Ann Westine Supercomputer and Workstation Communication Project Alan Katz produced a Supercomputer bibliography, and is beginning to document information he is finding out about supercomputers, and contemplating the design of a Virtual Windowing protocol for use with remote supercomputers. Command and Control Graphics Project No internet-related progress to report. Richard Bisbey Computer Center No internet-related progress to report. Seminars PAUL MOCKAPETRIS gave a seminar on the following papers: Chen, Peter Pin-Shan, "The Compact Disk ROM: How It Works," IEEE Spectrum, April 1986, pp.44-49. Oren, T. and G. A. Kildall, "The Compact Disk ROM: Applications Software", IEEE Spectrum, April 1986, pp.49-54. ALAN KATZ discussed the topic, "More about supercomputers and how people use them," (an update from the last seminar on this topic). Alan also reported on the paper: Smarr, Larry, "An Approach to Complexity: Numerical Computations," Science, 26 April 1985, page 403. GREG FINN discussed the topic: "Some Technical Problems in Metropolitan Scale Internetworking," Constructing a metropolitan scale internetwork will entail interconnecting potentially many thousands of component networks. Using today's internetworking technologies, problems are encountered which limit size and prevent uniform scaling. These weaknesses are discussed and suggestions are made regarding their solution. BRIAN HUNG gave a seminar on the paper: Green, Paul E., "Protocol Conversion", IEEE Transactions on Communications, Vol. Comm-34, No. 3, March 1986, pp.257-268. Ann Westine LINCOLN LAB .Grab=5; ----------- No internet-related progress to report. Jim Forgie LINKABIT .Grab=5; -------- .IOvr=3;1. I attended the Internet Engineering Task Force meeting on 8-9 April at BRL and presented suggestions on the possible evolution of EGP (RFC-975) and ideas stemming from my experiments on multiple-path routing algorithms (RFC-981). I also attended the Workshop on IS-IS Routing on 24 April at NBS and presented ideas on routing in the public-network community. 2. Mike Little has suspended work on his network-vulnerability study report due priority demands on another project. It does not seem possible to sustain the level of staff committment necessary to complete this effort, so it is being mothballed until happier times. 3. I refined the requirements document for Internet gateways and presented it for review and comment at several meetings sponsored by NSF, DARPA and NBS. Many useful technical suggestions were received and incorporated in the text, which will soon be submitted as an RFC. 4. Several unruly hosts began firing space junk across the DCnet Ether. These included a snatch of Martiangrams from net-127 hosts plus a retch of Atomicgrams (you first heard it here) to the Internet broadcast address (sic). Although at least the latter space travelers were destroyed over our DCN-GATEWAY spaceport, we have not been able to identify their port of embarcation, other than to verify the warp originated in the FORDnet sector. We did manage to find a Wollongong bug which resulted in broken CR sequences and bothered a bunch of hosts, including fuzzballs. 5. An examination of the congestion behavior of the DDCMP links used in FORDnet and DCnet and planned for use in the NSF Backbone net revealed occasional deadlocks in which each node in a ring of nodes waits for buffer space in the next downstream node. While the deadlock is eventually resolved by timeouts which reset the link protocol, the behavior can be debilitating and lead to connection timeout. I decided the time was ripe for some innovative, preemptive-queueing disciplines and began hacking up the fuzzware to that purpose. Initial results using a simple discipline that merely lops off the last customer on the longest queue were very encouraging. Further experiments based on time-space product, type-of-service and fancy fairness doctrines are close at hand. 6. Experimental type-of-service routing code was distributed to FORDnet, UMICHnet, UMDnet and DCnet fuzzballs and is in the process of being tested. The first experiment is to route interactive TELNET traffic via relatively low-speed and low-delay direct links between DCnet and UMDnet and bulk FTP traffic between these nets via high-speed and high-delay ARPANET/MILNET paths. The next experiment may be the same trick with the USAN satellite network just coming online now. .IOvr=0; Dave Mills MIT-LCS .Grab=5; ------- No internet related progress to report for April. Lixia Zhang NTA & NDRE .Grab=5; ---------- No report received. PROTEON .Grab=5; ------- The gateway engineering group moved this month, in several ways. First, the engineers in the group, who were scattered in free offices as they joined, moved into a continguous block of space that was freed for that purpose. Second, we moved from our trusty old PDP11 system to a new VAX system this month. (This means that we now have a compiler our compiler vendor will support; the PDP11 product is no longer supported.) The positive effects of all this have been fairly substantial! Also, as a measure of the increasing complexity and importance of the gateway effort, we have hired a software engineering manager to run the gateway project; this will release me to get back to work on more research oriented network work. I attended a meeting of the Internet Engineering Group and gave a presentation of some proposed new ICMP messages to meet some deficiencies in the interaction of hosts with the network switching layer; I will be writing this up as soon as I find the time (see above)! On the technical front, this month saw more work on a long-standing missing piece of the CGW; network up/down detection. Interfaces are periodically tested to make sure that they are still sending and receiving packets correctly; if not (and also on startup) a comprehensive self test is run. The protocol forwarding layers are notified when interfaces change state, and update their routing tables, etc, accordingly. As a by product of changes to the MOS I/O system to support this, basic I/O was sped up somewhat! Also this month, the prototype point-point synchronous interface finally came up, after a long struggle with the foibles of the Zilog SCC chip. A pair of these interfaces has been installed on the MIT-Proteon link for thorough workout, and has been performing excellently so far. Other sets are being sent out to customers as field demonstration units. Several new interface projects are ongoing; we hope to have results to report on them soon. Noel Chiappa SRI .Grab=5; --- .IOvr=4;1. Work continues on preparations for the upcoming network mobile host demonstration (F9) of the network reconstitution protocols. Because of the many changes in network topology, various problems in the handling of redirects in the Unix IP has surfaced in preliminary testing and are now being investigated. Essentially, the host will only accept and process REDIRECTs from gateways that conform to the host's (usually out-of-date) view of which nodes are up or down. 2. Mike Karels of U.C. Berkeley provided us guidance on UNIX 4.2 host implementation of our reconstitution protocol. We are thankful for his patience and hospitality. .IOvr=0; Jim Mathis As part of the transition of many SRI Sun Workstations to version 3.0 UNIX, we have been verifying, and converting as needed, the operation of our Multimedia Mail Software. At this point, we have completed the conversion of MPM, the mail transport module. Part of this effort included solving the poor performance of MPM on diskless Suns. This was traced to a problem with host lookup and yellow page operation. This has been fixed but may require some reordering of host tables for good performance on diskful systems. Verifying the operation of MMM, the mail user interface, and its suite of programs is currently underway. Preliminary results indicate that no changes will be required for 3.0 compatibility. On April 1, 1986 we participated with ISI and BBN in a multimedia conference using video, voice, and computer documents. Although the systems were not integrated during this initial experiment, much valuable experience user impressions were collected. As preparation for the discussion, a message composed of text, and images was sent to BBN using the SRI MMM system and integrated into Diamond. SRI is currently developing a MM conferencing prototype that will take realtime Navy-format messages and integrate them into a replicated database to be used as the information-base for the conference. We are currently implementing the module that willinterpret and transform the information from the Navy messages into graphics for the shared graphical workspace of the conference. Earl Craighill UCL .Grab=5; --- .IOvr=3;1. We have now established X.400 connectivity with the GIPSI system in INRIA (The French Institut pour Recherche en Informatique et Automatique), and we are exchanging messages on a regular basis. This includes gatewaying into RFC 822, with the address algorithms as specified in the ARPA/X.400 gatewaying spec (soon to be released as an RFC). 2. The UCL Butterfly gateway has arrived and is currently in our basement awaiting installation. 3. An IP-level access control scheme has been implemented in UCL's SATNET Access Machine in order to extend our control over internet access by UCL hosts. One reason for this has been the introduction of logical local nets consisting of student PCs on Ethernets. The SAM's route table indicates firstly whether or not access control is being enforced for particular local nets (a net may be a physical LAN or a logical local net). For those nets being controlled, IP datagrams passing through the SAM are checked to ensure that the UCL host concerned has the appropriate permission. 4. Work on upgrading our gateway service LSI-11s with 68000 boards is progressing. The 68K board now works with the DEQNA Ethernet driver. TCP, UDP and ARP are all working, while TFTP is nearly finished. 5. Kevin Miles (DEC, Reading, UK) visited UCL and held discussions with Robert Cole and Peter Lloyd about Transport protocol performance and the issues in migrating to ISO protocols. 6. Steve Wilbur attended the Privacy task force meeting at BBN (actually held at the end of March). .IOvr=0; Peter Lloyd TASK FORCE REPORTS .PBS; ------------------ APPLICATIONS .Grab=5; No report received. END-TO-END SERVICES .Grab=5; .IOvr=3;1. Steve Deering is making good progess on his initial implementation of an agent for IP multicasting, as described in RFC966. He has an initial version of the agent and corresponding host client code running in the V-System at Stanford. This code has also been shipped to BBN so they can contemplate its integration into the IP layer of 4.3BSD, to make it more widely available. 2. Also at Stanford, Dave Cheriton is working on VMTP, his proposed transaction protocol which will be efficient enough for use by distributed operating systems over an LAN. He has a V-Kernel version "limping" along, and says that the code has turned out to be "fast, relatively simple, and compact". He has prepared a paper on VMTP for a forthcoming conference. 3. The Task Force will be meeting at BBN early in May. .IOvr=0; Bob Braden INTERNET ARCHITECTURE .Grab=5; .IOvr=3;1. A third draft of a document on gateway requirements was circulated for advice and comment in INENG, INARC and the Workshop on IS-IS Routing held at NBS. Strictly speaking, this is an NSF document and is being reviewed both as a service to NBS and as possible guidance for the DoD community. 2. An agenda and marching orders are in place for the first meeting of INARC to be held at BBN on 8-9 May. Dave Clark has agreed to attend, as well as Lyman Chapin, who chairs the ANSI committee charged with routing issues. Marianne Gardner and possibly others from BBN will summarize their work on advanced routing issues. 3. Discussion continues over a proposed RFC on ISO addressing co-authored by Hans-Werner Braun and Ross Callon. This is a legacy of the now defunct GADS, but may serve as an opening wedge for more general Dod-ISO convergence issues. .IOvr=0; Dave Mills INTERNET ENGINEERING .Grab=5; .IOvr=3;1) The initial meeting of this Task Force convened during an open afternoon of the final GADS on January 16, 1986. A tentative Task Force agenda of short, mid and long term goals was set. 2) The first full meeting was held on April 8-9, 1986 where the agenda focused on: .IOvr=0; - Recent Internet performance degradation - EGP Modifications - IP refinements in hosts and gateways for improved routing and congestion control 3) A summary of recent Internet performance was given: .Grab=8; Dec 85 Jan 85 Traffic Sent by Mail Bridges ~27 ~35 (Mpackets/week) Traffic Sent by Mail Bridges ~90 ~105 (Mpackets/week) Traffic Dropped by Mail Bridges ~3% ~6% Traffic Dropped by Mail Bridges ~2% ~4% .IOvr=3;4) BBN discussed two possible sources for these sharp changes: 1) a bug in the LSI gateway routing software, which was recently discovered and corrected and 2) resource shortage in the Mail Bridge PSNs, due to be alleviated by end of April). 6) EGP modifications of two types were discussed: 1) areas in which the specification needed to be tightened and 2) areas in which the specification can profitable be `re-interpreted'. Examples of the second type include fragmented updates and interpreting the metric (eg, RFC975 - Autonomous Confederations). 7) Several new ICMP messages were proposed to facilitate fault isolation and routing (eg, initial gateway discovery, discovery of preferred address for multi-homed hosts). A co-operative congestion control scheme was proposed. Discussion of this scheme will continue in the Internet Architecture Task Force. .IOvr=0; 8) Actions in progress: - continued tracking of Internet traffic for expected improvements, - produce RFC of EGP modifications, - produce RFC specifying host attachment and IP/ICMP modifications. .IOvr=3;9) Detailed meeting notes are available upon request to corrigan@DDN1.ARPA, with cc to gross@MITRE.ARPA. Those requesting the notes will be added to the Task Force interest list. .IOvr=0; Phill Gross INTEROPERABILITY .Grab=5; No report received. PRIVACY .Grab=5; The Privacy Task Force welcomed a new member during the month of April: Steve Walker of Trusted Technologies, Inc. In response to a request, background material on ongoing task force discussions was distributed to John Laws of RSRE. Minutes of the March meeting are currently being prepared. John Linn ROBUSTNESS AND SURVIVABILITY .Grab=5; A meeting of the Robustness and Survivability Task Force was held at SRI-Washington on April 7th. We discussed the impacts of providing survivability features on new internetwork/network architectures. Jim Mathis SCIENTIFIC COMPUTING .Grab=5; No report received. SECURITY .Grab=5; No report received. TACTICAL INTERNET .Grab=5; The Tactical Internet task force is tentatively scheduled to meet on May 27, 1986 at MITRE in McLean VA. The agenda includes briefings on PR routing across unidirectional links and meteor burst demonstration activities in support of tactical intelligence communications. A detailed meeting agenda is in prepara- tion. Raj Mishra TESTING AND EVALUATION .Grab=5; No report received.