[IMR] IMR88-02.TXT FEBRUARY 1988 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 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 network mail to Ann Westine (Westine@ISI.EDU) or Karen Roubicek (Roubicek@NNSC.NSF.NET). BBN LABORATORIES AND BBN COMMUNICATIONS CORPORATION --------------------------------------------------- WIDEBAND NETWORK A new BSAT release has been distributed to the Wideband Network sites which corrects a number of bugs that have been identified in the BSAT's processor node and satellite channel I/O device software. The occurrence of sporadic BSAT system-level restarts and glitches in channel connectivity has been dramatically reduced since the distribution of this release. Westine [Page 1] Internet Monthly Report February 1988 The SATNET and the Wideband Network were directly connected for the first time this month via a single Butterfly Internet Gateway located at DCEC in Reston, VA. With the addition of the Wideband Network interface, the DCEC gateway now has four network legs: the ARPANET and EDN net, as well as the SATNET and the Wideband Network. Direct SATNET-Wideband connectivity has so far been sporadic; the source of this problem is being investigated. The provision of a stable SATNET-Wideband connection may result in improved access to certain West coast hosts for European Internet users. The Wideband Network provided the wide-area connectivity for a successful three-site demonstration of the Cronus Distributed Operating System on February 17. The demonstration was given as part of USAF/ESD's Technology Validation Experiment (TVE). It incorporated participating Cronus "clusters" located at RADC, ESD/Mitre, and BBN. SATNET February has been a busy month for the SATNET. A new modem was shipped to Tanum early in the month to replace the channel 1 modem that had gone bad. Tanum was then back in operation on both channels. At about that time, a higher than expected error rate was observed on both channels. Coordinating with the four earth stations, the power levels were adjusted and the error rate returned to normal. We have distributed a new software patch to the SIMPs that increases the throughput of the SATNET when two channels are in operation. The Measurement Task Force is continuing work to characterize the IP and TCP behavior of the SATNET. Late in the month, we experienced an outage on the CSS to Roaring Creek phone line. The DCEC to Roaring Creek phone line was also down at the time isolating the SATNET from the ARPAnet. When the CSS line was repaired, the Roaring Creek PSP terminal and the channel 0 uplink at Tanum were found to be experiencing problems. The problem at Roaring Creek has been fixed but we are still working on the problems at Tanum. The DCEC to Roaring Creek phone line was repaired on Tuesday, March 1. INTERNET R&D A new Butterfly Gateway was installed in Ottawa, Canada. This gateway replaces the LSI-11 gateway there and is the first Butterfly Gateway on the Arpanet with X.25. The installation went well and the gateway has been stable since. A new interface was added to the DCEC gateway, which provides a direct connection between the Satnet and the Wideband network. Westine [Page 2] Internet Monthly Report February 1988 This will have the affect that traffic to/from Satnet which goes to destinations directly connected to the Wideband network, will not go over the Arpanet. We submitted several documents for the Internet Engineering Task Force IDEA series. They are: o IDEA002.TXT - A Comparison of "Link State" and "Distance Vector" Routing Algorithms, Ross Callon (BBN) o IDEA003.TXT - Guidelines for the use of Internet-IP addresses in the ISO Connectionless-Mode Network Protocol, Ross Callon (BBN) and Hans-Werner Braun (U Mich) o IDEA007.TXT - Requirements for Inter-Autonomous Systems Routing, edited by Ross Callon (BBN) for the Open Routing Working Group o IDEA009.TXT - Exterior Gateway Protocol (EGP), Version 3, Marianne Gardner (BBN) and Mike Karels (UCBerkeley) Bob Hinden (Hinden@BBN.COM) ISI --- Internet Concepts Project Jon Postel and Paul Mockapetris visited Hanan Potash at the Scientific Computing Center in San Diego, Feb 17. Jon Postel, Danny Cohen, and Walter Prue hosted the Los Nettos meeting at ISI, Feb 1. Steve Casner hosted the UITF meeting at ISI, Feb 10-11. Eight RFC's were published this month. RFC 1041: Rekhter, J., "Telnet 3270 Regime Option", T.J. Watson Research Center, IBM, January, 1988. RFC 1042: Postel, J., and J.K. Reynolds, "A Standard for the Transmission of IP Datagrams over IEEE 802 Networks", USC/ISI, February, 1988. RFC 1043: Yasuda, A., and T. Thompson, "TELNET Data Entry Terminal Option DODIIS Implementation", Defense Intelligence Agency, February, 1988. RFC 1044: Hardwick, K., and K. Hardwick, "Internet Protocol on Network Systems HYPERchannel Protocol Specification", NSC, and NASA-Ames GE, Westine [Page 3] Internet Monthly Report February 1988 February, 1988. RFC 1045: Cheriton, D., "VMTP: Versatile Message Transaction Protocol, Protocol Specification", Stanford University, February, 1988. RFC 1046: Prue, W., and J. Postel, "A Queuing Algorithm to Provide Type-of-Service for IP Links", USC/ISI, February, 1988. RFC 1047: Partridge, C., "Duplicate Messages and SMTP", CIC at BBN Labs, February, 1988. RFC 1048: Prindeville, P., "BOOTP Vendor Information Extensions", McGill University, February, 1988. Multimedia Conferencing Project We finally received the Image 30 video codec boards from Concept Communications this month. The boards have been checked out in a self-loop mode, and we have begun modifications to the on-board firmware to implement HDLC serial line protocol required to interface to the Butterfly. Coding of changes to accommodate the new codec have been completed for the packet video program (PVP) that runs in the Butterfly. The major difference between processing for the Image 30 and the old ISI codec is that the frame rate is constant for the Image 30 and that the segments of each video frame must be delivered in order and uninterrupted. (Damaged frames may be discarded without loss of synchronization.) Check-out of this code awaits completion of the codec firmware mods. The Image 30 codec has low delay and a fast frame rate (30/sec) compared to other codecs, but its resolution is lower and there are some unusual motion effects. In the future we plan to explore variable frame rate and/or data rate to try to minimize the motion effects. We may also be able to improve resolution by trading off some of the frame rate. The coding in PVP was done with this evolution in mind. Dave Walden and Steve Casner (DJWalden@ISI.EDU, Casner@ISI.EDU) Brian continues working on the simulation of the Echo Canceller Algorithm. Brian decided that this would be worthwhile because no information was available which indicates how the system was going to behave for different values of Loop Gain, i.e., for what the value of K will the system become stable. Brian was also interested in how the Westine [Page 4] Internet Monthly Report February 1988 system would behave for different kinds of inputs and attenuation through the Hybrid and the value of the 64 tap weights. The simulations were done with different inputs, amplitudes, loop gain, and attenuation. The results were very useful in pointing out the value of loop gain necessary for a stable system and also for a particular convergence rate. Brian Hung (Hung@ISI.EDU) NSFNET Project Bob Braden and Annette DeSchon completed a major revision to the NNStat statistics package, Release 2.0. This release fixed a number of bugs that created reliability problems in earlier versions. It also included significant extensions in functionality. The remote command interface to statspy was entirely rewritten, and remote attach commands were implemented. Complete remote control of statspy is now possible using the rspy program. A new data reduction program in R2.0 will look up all IP addresses in recorded data files and add host/network names. We plan to release NNStat source very shortly. Bob has begun a paper describing the internal organization and algorithms used by statspy. Bob Braden attended a meeting of the Federation at Ann Arbor, which was largely concerned with the new NSFnet backbone. He also participated, as an IAB member, in a one-day review of the competing network management protocols. Bob Braden (Braden@ISI.EDU) Supercomputer and Workstation Communication Project Alan wrote a graphics display program which runs under X- Windows and a front end to it which runs in GNU Emacs Lisp. This allows Alan to do graphics programming in the Emacs Lisp environment. Alan has also been experimenting with some Mandlebrot set graphics using this program. Alan continues work on a split editor for GNU Emacs, and has written more utilities that run under X-Windows including a screen snapshot program. Alan Katz (Katz@ISI.EDU) Westine [Page 5] Internet Monthly Report February 1988 MIT-LCS ------- The network simulator we are building has reached another stage in maturity. Based on an IP network substrate, it can now emulate 3 versions of pseudo Berkeley UNIX TCP implementations: 4.2 (which retransmits all after a timeout), 4.3 (which retransmits only one segment after a timeout), and slow-start algorithms as proposed by Van Jacobson. New graphic tools are also available. In addition to graphic displays of the dynamic load at each point in the network (such as channel busy status, packet queue length at gateways), the simulator now can produce graphic prints of simulation results, such as the curve of TCP transmission sequence number versus time. To experiment with rate-based flow control mechanism, a pseudo version of NETBLT protocol has been installed into the simulator. Lixia Zhang (Lixia@XX.LCS.MIT.EDU) MITRE Corporation ----------------- No report received. NTA & NDRE ---------- No report received. SRI --- No internet-related progress to report. Zaw-Sing Su (ZSu@SRI.COM) UCL --- No internet-related progress to report. John Crowcroft (jon@CS.UCL.AC.UK) Westine [Page 6] Internet Monthly Report February 1988 UNIVERSITY OF DELAWARE ---------------------- 1. Traffic on the NSFNET Backbone continued to escalate throughout the month, with over 120 nets now on the backbone and approaching 400 nets on the core system. Some sites are reporting increasing incidence of temporary lockup conditions with the DEQNA Ethernet interface and/or driver, which is believed due to escalating traffic. Minor changes to the preemption strategy were made in order to help relieve the pressure and to resist route-flaps due gratuitous redirects. 2. Work continued on robust clock-synchronization algorithms to improve accuracy, decrease vulnerability, reduce network overheads and support vast hordes of chimers. The fuzzball NTP implementation was again modified to support a variable-rate polling feature in which the polling interval is dynamically adjusted to varying conditions of network delay dispersion and loss rate. On most paths this allows the polling interval to be increased to as much as 1024 seconds (17 minutes) once initial synchronization has been achieved, but without significant loss in accuracy or robustness. 3. New software versions including the latest NTP support were distributed to all fuzzballs in the US and Europe. A configuration for a primary service network including five multiply-redundant time servers at ISI, NCAR, U Maryland, U Delaware and Ford Research was designed, implemented and is now in operation. It appears that DECWRL, MIT and NASA/AMES may join this network, which is designed to tick wholesale time to secondary time servers at local-net confluences, which then tock to the retail customers, perhaps using other time protocols such as Unix 4.3bsd timed. 4. A document "Network Time Protocol (Version 1): Specification and Implementation" is now in final draft and ready for RFC submission. Mike Petry reports a companion NTP daemon for Unix 4.3bsd is nearing test status. 5. Mike Minnich is now wickedly abusing a pair of Proteon gateways in order to assess their overload and fairness behavior. A raft of GFE stuff, including a Sun workstation, WWVB clock and other trappings was finally prized loose from another place and installed in my lab, which now boasts three Suns, two PCs, four fuzzballs, and rich connectivity to campus nets, NSFNET and ARPANET. Dave Mills (Mills@UDEL.EDU) Westine [Page 7] Internet Monthly Report February 1988 NSF NETWORKING -------------- UCAR/BBN LABS NNSC The NNSC published and distributed the second NNSC bulletin, and the third issue of the NSF Network News was completed for distribution in March. Please send requests for these publications to . Karen Roubicek attended the meetings of the Federation of American Research Networks hosted by MERIT in Ann Arbor, and Craig Partridge participated in the Network Management Review Meeting held in San Diego as well as the USENIX meeting in Dallas. By Karen Roubicek (roubicek@nnsc.nsf.net) NEW NSFNET BACKBONE At the recent Federation Meeting held in Ann Arbor on the 8th and 9th February, we briefed the Federation of American Research Networks about the new NSFNET backbone. During the informational sessions, various related technical issues were presented and discussed. This was the very first direct communication between Merit/IBM/MCI and the sites for the new NSFNET backbone. After the Federation meeting Hans-Werner Braun, Scott Brim, Ed Krol, Jim Ellis, Gene Hastings, and Karen Roubicek met informally to discuss some of the issues with the transition from the current NSFNET backbone to the new one. They have outlined some scenarios and will have followup discussions to make the transition smoothly and as transparent to users as possible. The communication links for the four node test network located in Ann Arbor MI, Reston VA, Milford CT and Yorktown NY are installed and connected to the IDNX equipment. Work is underway to make this network operational. The IBM 4381 system for the Information Services for the new NSFNET backbone has been installed and the installation of software is in progress. We anticipate that the Information Services system will begin to be available before the operational network becomes functional. The IS system is currently connected to a Merit Ethernet and supports Telnet, FTP and SMTP. We are continuing to beta test the Wisconsin ARGO (OSI) code. We were able to test the ARGO implementation between the University of Michigan and the University of Wisconsin via non-encapsulated ISO packets. This is possible via the bridged USAN satellite network which reaches the University of Wisconsin within two satellite hops (which means approximately a one second round trip delay). The Westine [Page 8] Internet Monthly Report February 1988 testing was done largely by Rob Hagens at the Wisconsin site and Bilal Chinoy at Michigan. So far we have tested a Ping equivalent (CLNPPING) and remote login (Rlogin) between the two sites. Both looked almost indistinguishable from their TCP/IP equivalents. There are some remaining problems with the TP4 code, as the version we have was optimized for local low delay data networks. Those problems were anticipated and, in fact, part of the reason for the USAN test, which allows for high delay testing, but with predictable delays. The kernel version at Wisconsin should allow for modulating the TP4 characteristics to overcome these problems. We are testing a version of "gated" modified by Jeff Honig (Cornell) to allow for EGP communications between a RT/PC and a Proteon gateway. Further testing is planned on different vendors' gateways. This testing is necessary to verify EGP communication capability between the new NSFNET backbone and the attached regional networks. We also did some testing between a Unix gated and a Fuzzball in case EGP may need to be used at some of the current NSFNET sites. By Jessica Yu (jessica_yu@um.cc.umich.edu) NSFNET BACKBONE SITES CORNELL UNIVERSITY THEORY CENTER No report received. UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN/NCSANET The connection from uxc.cso.uiuc.edu to IMP 94 port 6 is up and running. UIUC is currently draining NSF traffic from Midnet and the Chicago area to the ARPANET. We will take on more of the NSFnet traffic when local routing becomes a bit more stable and folks get back from IETF. Chicago area network is progressing smoothly. The T1 between Urbana and Chicago has been up since early January. University of Chicago should be connected to the Chicago end via T1 within the next two weeks. Northwestern should be soon after that. Notre Dame is scheduled for early April via 56Kbps. Fermi Lab is installing a 14.2Kbps line (TCP/IP only) to Urbana. We are hoping to strike a deal whereby a 56Kbps line goes in between the Chicago end of the T1 and Fermi. This could provide HEPnet/SPAN connectivity for all of the Chicago area network participants as well as UIUC/NCSA. (Currently all participants run 9.6Kbps lines to Fermi strictly for this purpose.) A single shared line would save hundreds of dollars monthly. Westine [Page 9] Internet Monthly Report February 1988 No schedule is set for Argonne Nat'l Lab to upgrade to T1 or for Illinois Institute of Technology to connect via 56Kbps. By Charlie Catlett (catlett@ncsa.uiuc.edu) and Ed Krol (krol@uxc.cso.uiuc.edu) JOHN VON NEUMANN NATIONAL SUPERCOMPUTER CENTER No report received. NATIONAL CENTER FOR ATMOSPHERIC RESEARCH & UNIVERSITY SATELLITE NETWORK PROJECT Intermittent transmitter radio problems have at the NCAR site have plagued USAN this month. Radio swaps have been made and it is hoped that the problems will not reoccur. The satellite connection on USAN of the Naval Research Lab in the Washington D.C. area is in place and checked out. Actual packets flowing on this link are still awaiting the ethernet connection at NRL. By Don Morris (morris@scdsw1.ucar.edu) PITTSBURGH SUPERCOMPUTING CENTER No report received. SAN DIEGO SUPERCOMPUTER CENTER We were visited by a Tiger Team from CRC to check on our PSN. They decided that our's was improperly configured; they requested replacement hardware. Since this will take at least 60 days, we are still at least that far from having an operational connection (trunk lines are still pending, of course). Our Proteon p4200 has had a pretty good month - some "memory loss" on routes which a reboot cleared. As the problem has not reappeared, we are just keeping an eye on it. We have installed the latest version of SRI's MultiNet on the Center's VAX's. Finally, the FTP server for our Cray mentioned last month, is now in production use. Again, thanks to NCSA on which this server is based. By Paul Love (loveep@sds.sdsc.edu) Westine [Page 10] Internet Monthly Report February 1988 NSFNET MID-LEVEL NETWORKS BARRNET (No report received) JVNCNET (Refer to JVNNSC backbone report) MERIT/UMNET The PSCNET T1 connection at the University of Michigan was established this month and now provides for our primary path to the Internet. The average delay times between the University of Michigan and PSC-GW dropped from about 700 msec to about 35 msec. About 550 msec of our previous connection were due to the USAN satellite delay. Our traffic to PSC-GW now traverses three Proteon gateways. New tools are available for the Merit PCPs and SCPs to collect packet statistics for IP traffic carried by the network nodes, as well as to display the IP routing table at the nodes. Work will start soon on implementing TCP within the Merit packet switching nodes. This will be augmented by a Telnet implementation shortly thereafter to allow direct terminal access to Telnet servers for the currently almost 9500 terminal ports within the Merit network. By Jessica Yu (Jessica_Yu@um.cc.umich.edu) MIDNET MIDnet did not have any significant problems during the month of February. UNL took our road show to 3 MIDnet members during the month. We found this to be a very valuable experience from both our point of view and the local campus point of view. We learned a grea deal about how they are trying to connect their campuses to MIDnet, the problems they are experiencing and the sort of assistance that will be of most use to them. The persons on the campuses learned some about the nature of the problems that they are having with connections to various NSF sites and ways that they can diagnose and solve those problems. We will be continuing these trips in March and April and hope that the rest will have as good an effect as the previous ones. By Dale Finkelson (dmf@fergvax.unl.edu) MRNET The MRNet link to NSFNet via NCSA is working well. A local hack running on an NCSA systems allows hello routing information to get to the uc.msc.umn.edu gated system on MRNet via the Proteon 4200's Westine [Page 11] Internet Monthly Report February 1988 that link MRNet to NCSA. (ie, not via rip) Mayo Clinic is now up and is 129.176.0.0 By Thomas Jacobson (thomas@umn-rei-uc.arpa) NCSANET (Refer to the UIUC & NCSAnet backbone report) NORTHWESTNET NorthWestNet is now fully connected and is using two routes to NSFnet--via 56 kbps satellite to NCAR from Oregon State University and via 56 kbps land line from University of Washington to SDSC. The Boeing Companies have made a renewable grant of $250,000 worth of CRAY cycles to NorthWestNet, and Management Committee has established process for distributing the grant. The Training Committee is negotiating with University of Washington to coordinate NorthWestNet training activities. John Skelton, Oregon State University, is the newly elected Chairman of the Technical Committee. The Network has requests from two additional institutions for connection. By Dick Markwood (MARKWOOD@vaxf.colorado.edu) Westine [Page 12] Internet Monthly Report February 1988 NYSERNET As of 1 January 1988, NYSERNet had the following topology: Clarkson Syracuse--+ | NISC | | | Rochester--------Cornell---------RPI---Albany | || || | ....Alfred || || Buffalo...Fredonia || || | ....Oswego || || | || || Binghamton || +-------- || ------StonyBrook | || | || | || | || CUNY--NYTEL/NSMAC--Columbia======NYU==Rockefeller | |\ | | /| | || | | \ | NYNEX/S&T / | | || | | \ BNL / | | WP/CO | | \ / | | | | +-------------POLY---+ | | | | | | | +------------------------------+ | | | +--------------------------------------+ || ==== || T1 || \ | / 56kbits .... 9.6kbits The Rockefeller/NYU link was upgraded from 56kbits to T1 in February The NYTEL White Plains Central Office (WP/CO) was connected by T1 to the Rockefeller node, in preperation for connection to IBM Yorktown Heights. The RPI/NYU link was upgraded from 56kbits to T1 on March 1st. Alfred University was connected to NYSERNet in February at 9.6kbits. NYSERNet's SGMP/SNMP V2.0 (a NOC implementation) was announced for licensing to the INTERNET community. Representatives of NYSERNet participated in the Federation/Merit meeting in Ann Arbor. Mark Fedor and I participated in the "Network Management Convergence Meeting" on the 29th of February in San Diego. Westine [Page 13] Internet Monthly Report February 1988 By Marty Schoffstall (schoff@nisc.nyser.net) SDSCNET (Refer to SDSC backbone report) SESQUINET The complete initially proposed SesquiNet configuration has been operational now for several months. In addition, we have added one campus this month (Prairie View A&M). The following campus networks are being served, and are advertised via EGP to the core: Baylor College of Medicine 128.249 Houston Area Research Center 192.31.87 Prairie View A&M University 129.208 Rice University 128.42 Texas A&M University 128.194 Texas Southern University 192.31.101 and the University of Houston 129.7 The serial line from NSFnet/NCAR to SesquiNet/Rice has been operational for several months, and routes to SesquiNet via NSFnet are now being advertised. Performance is very good. We have fixed one source of occasional outages (a marginal local loop in Houston), and have enjoyed good stability. We are still testing cisco's support for dual protocol (IP and DECnet) routing. We are also testing cisco's support for 1822-HDH ARPANET support. This should be operational during the month. By Guy Almes (almes@rice.edu) SURANET No report received. Westine [Page 14] Internet Monthly Report February 1988 WESTNET 1. The paperwork to proceed with hiring a full time technical support person is proceeding through our Office of Affirmative Action. We expect to have this person on board in April, 1988. 2. We have met with MCI, who was the low bidder for our inter- LATA circuits, to establish formal trouble reporting procedures and centralized lines of reporting. 3. The HELLO Protocol experiment has been proceeding well. 4. We are still waiting for MCI and Mountain Bell to install the remainder of our circuits. We hope that most of them will be in place by March 30, but have been promised that they will be in place no later than April 20, 1988. By Pat Burns (pburns%csugreen.bitnet@CUNYVM.CUNY.EDU) Westine [Page 15] Internet Monthly Report February 1988 TASK FORCE REPORTS ------------------ APPLICATIONS -- USER INTERFACE No report received. AUTONOMOUS NETWORKS No internet monthly progress to report. Deborah Estrin (Estrin@OBERON.USC.EDU) END-TO-END SERVICES No progress to report. Bob Braden (Braden@ISI.EDU) INTERNET ARCHITECTURE The most recent meeting of the Internet Architecture Task Force was held at BBN on 17-18 December 1986. It was organized as a workshop, with the primary focus on evolution of the next-generation internet, whether based on the existing Internet or not. The goals of the workshop were intended to foster exploration and evaluation of new technology specifically required for the very large, very fast internet of the future, with special relevance to research networks under consideration by DARPA, NSF and collaborating agencies. Presentations were prepared, but informal, and there was opportunity for discussion and debate. While most presentations did not involve a formal paper, several did include briefing aids and other archival material, which was distributed to the attendees at the meeting. For additional copies, please contact the individual authors. Speaking for myself and I believe most of the attendees, the meeting was successful, interesting and brought together members from diverse research communities not captured in one room since the days of the old Internet Research Group meetings. It is clear that the energy level of the community is healthy and sustainable; however, it is also clear that focus and support are needed to channel this energy toward technical realization, demonstration and evaluation. Westine [Page 16] Internet Monthly Report February 1988 The following summary of the workshop was edited by Mike Minnich from material submitted by the various session chairs. My sincere thanks go to the participants, session chairs and especially Mike. Thursday, December 17 MORNING SESSION 1. Standards, Evolution, and Requirements CHAIR: Paul Tsuchiya, MITRE Talk: "User Requirements", Miles Fidelman, BBN Talk: "User Requirements, DOE Research", Bill Johnston, LBL Talk: "Standards and Evolution", Paul Tsuchiya, MITRE 2. Very High Speed Networks and Switching CHAIR: Jon Turner, Washington University Talk: "Key Issues for Public Networks", Jon Turner, Wash. University Talk: "BPN Architecture", Guru Parulkar, Wash. University Talk: "Very High Speed Networks", Jil Westcott, BBN AFTERNOON SESSION 3. New Architectures, other than Very High Speed CHAIR: Dave Clark, MIT Talk: "Next Generation Internet Protocol", Ross Callon, BBN Talk: "The Network Interface Buffer Architecture", Dave Clark, MIT CHAIRPERSON'S COMMENTS: This session related to those architectural issues not specifically related to high speed. The first talk, by Ross Callon of BBN, described requirements and approaches for a next generation Internet. His conclusions are: 1. Policy controls on allocation are needed, for example congestion control. 2. Routing is needed that does not require complete routing tables, for example partial source routes. 3. We need more flow state. 4. The IP datagram model will not be sufficient. The next talk was by David Clark of MIT LCS, and described a host-network interface architecture which would permit a high-performance division of Westine [Page 17] Internet Monthly Report February 1988 networking function between the host and a separate outboard processor. The major component of this interface architecture was named messages, where the names are used by the controller to map the message to the intended buffer. The following discussion considered the extent to which bandwidth need be considered a critically expensive resource in the next generation network. As the net provides a more complex set of types of service, it may be able to load the links more heavily while still meeting the requirements of the users. However, the techniques to do this are not clear, and the needed speeds of switching may not permit support of multiple types of service. There was no consistent conclusion. Friday, December 18 MORNING SESSION 4. Naming and Directory Services CHAIR: Debbie Deutsch, BBN Talk: "Addressing, Routing, Mobile Hosts", Keith McCloghrie, TWG Talk: "Portable IP Hosts", Jose Rodriguez, Unisys Talk: "Naming and Directory Services", Debbie Deutsch, BBN CHAIRPERSON'S COMMENTS: Three presentations were given, two concerned with addressing at the Internet level, and one with named for users and applications. Keith McCloghrie from Wollongong, made a proposal for logical addressing in the Internet. Packet addresses, perhaps the same as existing IP addresses, would identify the destination of a packet. Separate routing addresses would be used for routing. The conversion from packet addresses to routing addresses would be performed by address servers. The packet addreses would be relatively stable while routing addresses could change with time. While routing based on routing addresses would be hierarchical, "back-door" routes, based on unilateral aggreements, could circumvent potential inefficiencies. Jose Rodriguez discussed the Portable Host Access Component Project at Unisys. The goal of this project is to provide DDN connectivity to portable personal computers as hosts. Doing this requires some flexibility in the way Internet addresses are used. The Internet architecture needs a way to deal with such mobile hosts. Two potential approaches are shadow hosts, much like IP multicast agents, and real-time name servers. The last presentation, which was given by Debbie Deutsch of BBN Labs, concerned naming and directories for applications and their users. There currently is no framework for naming and managing Westine [Page 18] Internet Monthly Report February 1988 these entities in the Internet. The current work of ISO and the CCITT (the X.500 series) can provide a basis for research in this area and its applications to distributed systems, remote services, and management at the application level. 5. Engineering Issues CHAIR: Len Bosack, cisco Talk: "Merit/IBM/MCI Proposal for NSFNET", Jacob Rekhter, IBM Talk: "Dissimilar Gateway Protocol", Mike Little, M/A-COM Linkabit Talk: "Internetting: DARPA Concerns", Mark Pullen, DARPA AFTERNOON 5. Engineering Issues, continued Talk: "Inter-AS Routing", Ross Callon, BBN Talk: "Policy Based Routing", Tassos Nakassis, NBS 5a. Bomb Scare 6. General Services CHAIR: Miles Fidelman, BBN Talk: "DOE Energy Sciences Networking", Bill Johnston, LBL Talk: "Multi-media Conferencing", Claudio Topolcic, BBN CHAIRPERSON'S COMMENTS: The most notable emerging service requirement is for much higher bandwidth (megabit to terrabit). Bandwidth requirements are increasing both in the aggregate, and for support of individual applications: 1. In the short term, new load is stemming primarily from increasing numbers of host connections, in particular mainframe systems on the military side of the net, and PCs all over the place. These connections are leading to an increasing aggregate requirement but not increasing the bandwidth required by individual application sessions. 2. It is perceieved that the continued increase in available computing power (e.g. gigaflop processors) will lead to requirements for equivalent growth in communications support (e.g. gigabit and terrabit networks). Discussions about current support requirements for supercomputer centers does not seem to support this case - since the primary applications seem to be remote job entry and the return of results in files comparable in size to current files routinely passed across the Internet (though this does seem to indicate that there will be specific hosts sites that Westine [Page 19] Internet Monthly Report February 1988 are the locus for large volumes of traffic). 3. In contrast, the distribution of modest amounts of computing power around the network, coupled with communications intensive applications (e.g. multi-media conferencing, distribution of large technical data packages) do seem to require that significantly higher bandwidths be delivered to end applications. Estimated requirements for multi-media conferencing are on the order of 1-100 Mbits/Second, a technical data package for a large system can easily fill an optical disk (i.e. gigabytes of data). Also notable is an increasing requirement for selectable qualities of service. For example, real-time multi-media conferencing requires end-to-end delays in the .1 second range, while distribution of large files does not. An increasing sensitivity to cost also argues for quality of service mechanisms that select on cost (coupled with chargeback mechanisms). The increasing use of the Internet to support researchers (as opposed to research) is leading to a need to simplify the operational characteristics of the Internet. On the military side, there appears to be an intent to assign responsibility for data communications to the telephone organizations in place on military bases. Clearly the current problems associated with managing gateways have to be cleared up. The increasing connection of large mainframes to the Internet will require that network management mechanisms be interfaced to host-based management software that expects to be able to see underlying networks. Finally, it appears that the powers that be are moving toward charging for usage - accounting mechanisms will need to be embedded in the Internet. 7. Security and Privacy Issues CHAIR: Mike Padlipsky, MITRE Talk: "An OSI Approach to Network Security", Mike Padlipsky, MITRE CHAIRPERSON'S COMMENTS: The Security session, only added to the agenda during the first morning, consisted of a single presentation, by its ad hoc Chairman, Michael Padlipsky. An attempt to distill and abstract several years' worth of thinking and work on an Outboard Processing Environment (formerly known as Network Front-End) based security architecture, the talk was deliberately misleadingly titled "AN OSI Approach to Network Security". An overlay turned the OSI into Orthogonal Security Interposition, by which was meant in essence that the recommended approach would consciously avoid being intermingled with the conventional protocol suite(s) in play (the "orthogonality") and would operate by being "interposed" between Westine [Page 20] Internet Monthly Report February 1988 the Host and its offloaded protocol interpreters, interecepting the transactions of an appropriately designed Host-Front End protocol and taking appropriate actions for labeling, auditing, and encryption in concert with counterpart implementations of the security mechanism at the target system. It should be noted that it was primarily the fact that Dave Clark had actually spoken in favor of protocol offloading the day before which inspired the speaker to be willing to risk exposing these thoughts in his presence, but the approach has long been believed to have merit in those environments (DoD, e.g.) where the necessary ubiquity of the security mechanism-bearing OPEs can be mandated, thus justifying the amount of "assurance" one would need to posit for the OPEs. (Indeed, assurance considerations are even more important as a reason for the recommended orthogonality than the also interesting capablility of avoiding being bound to a potentially changing protocol suite which could be conferred by an H-FP of the RFC 928/9 sort.) Dave Mills (Mills@HUEY.UDEL.EDU) INTERNET ENGINEERING The planned agenda for the March 1-3 IETF meeting in San Diego is below. A more detailed report on the meeting will be given in the March report. There have been 9 documents produced by the IETF since the last meeting (IDEAS 002-010). Many of these are planned to be submitted as RFCs. A major activity for the various working groups at the March 1-3 meeting will be to discuss these documents, incorporate comments and perform any final editing to prepare them for submittal (or for the next level of discussion). As a reminder, the papers are: o IDEA002.TXT - A Comparison of "Link State" and "Distance Vector" Routing Algorithms, Ross Callon (BBN) o IDEA003.TXT - Guidelines for the use of Internet-IP addresses in the ISO Connectionless-Mode Network Protocol, Ross Callon (BBN) and Hans-Werner Braun (U Mich) o IDEA004.TXT - Routing Information Protocol, Chuck Hedrick (Rutgers) o IDEA005.TXT - Requirements for an Open Internal Gateway Protocol, edited by John Moy (Proteon) for the Open IGP Working Group o IDEA006.TXT - ISO Presentation Services on top of TCP/IP-based Internets, Marshall Rose (TWG) Westine [Page 21] Internet Monthly Report February 1988 o IDEA007.TXT - Requirements for Inter-Autonomous Systems Routing, edited by Ross Callon (BBN) for the Open Routing Working Group o IDEA008.TXT - The Responsible Person Resource Record, edited by Louis Mamakos (U of Md) for the Name Domain Working Group o IDEA009.TXT - EGP3, Marianne Gardner (BBN) o IDEA010.TXT - The DCA Protocol Testing Lab, Unisys. 003 and 006 will be discussed in the OSI WG; 004 in the Short-term Routing WG; 005 in the Open IGP Routing WG; 007 in the Open Routing WG; 008 in the Domain WG and 009 in the EGP3 WG. Ross Callon will be presenting aspects of 002 and other topics on Thursday. All IDEAS are available by anonymous ftp from the directory on SRI-NIC.ARPA. There is a brief synopsis of each IDEA in IDEA-INDEX.TXT on SRI-NIC.ARPA. Agenda for the March 1-3 IETF ------ --- --- ----- --- ---- TUESDAY, March 1 8:30 am Opening Plenary (Introductions and local arrangements) 8:45 am Working Group meetings convene - Open Systems Routing (Hinden, BBN) - Open Systems Internet Operations Center (Case, RPI) - Authentication (Schoffstall, RPI) - Open IGP (Petry, UMD/Moy, Proteon) - Internet Host Requirements (Gross, Mitre/Braden, ISI) 5:00 pm Recess WEDNESDAY, March 2 8:30 am Opening Plenary 8:45 am Working Group meetings convene - Short-Term Routing (Hedrick, Rutgers) - Domains (Louis Mamakos, UMd) - Performance and Congestion Control (TBA) - EGP3 (Gardner, BBN) - OSI Technical Issues (Rose, TWG) 11:30 am Lunch 1:00 pm Detailed Report on the New NSFnet (Braun, UMich/Rechter, IBM) - including overview, topology, network management, packet switch architecture and routing. 3:00 pm Break 3:15 pm Status of the Adopt-a-GW Program (Enger, Contel/Gross, Mitre) - Remember the pre-Christmas `brain-dead' message? Donors of equipment will be credited and data will Westine [Page 22] Internet Monthly Report February 1988 be presented showing the improvements. 3:45 pm BBN Report (Hinden/Gardner, BBN) - Background and status of the Core gateway upgrade to butterflies; Internet and network operations report, including the PSN 7.0 and new EE transition. 5:00 pm Recess THURSDAY, March 3 8:30 am Opening Plenary 8:45 am Working Group Reports and Discussion - OSI Technical Issues (Rose, TWG) - Open Systems Routing (Hinden, BBN) - Short Term Routing (Hedrick, Rutgers) - Open Systems Internet Operations Center (Case, RPI) - Authentication (Schoffstall, RPI) - Open IGP (Petry, UMD/Moy, Proteon) - Internet Host Requirements (Gross, Mitre/Braden, ISI) - Domains (Louis Mamakos, UMd) - Performance and Congestion Control (TBA) - EGP3 (Gardner, BBN) - InterNICs (Feinler, SRI) 11:30 am Lunch 1:00 pm Technical Presentations - Routing and Miscellanea (Callon, BBN) - Internet Multicast and VMTP (Deering, Stanford) - TCP Performance Prototyping (Van Jacobson, LBL) 3:00 pm Break - Cray TCP Performance (Borman, Cray Research) - DCA Protocol Testing Laboratory (Swanson, Unisys) 4:45 pm Concluding Plenary Remarks 5:00 pm Adjourn Phil Gross (Gross@MITRE.GATEWAY.ORG) INTERNET MANAGEMENT No report received. PRIVACY The IAB Privacy Task Force had a productive two-day meeting at Lawrence Livermore National Laboratory in Livermore, California on 2 and 3 March. Attendees were: Dave Balenson, Curt Barker, Don Brinkley, Morrie Gasser, Steve Kent, John Linn, Dan Nessett, and Steve Wilbur. Most of the meeting's discussion could be divided into two basic categories, mail-related and mail-unrelated. Mail- related discussion included a development status review, Westine [Page 23] Internet Monthly Report February 1988 discussion of RFC-1040 and on comments received thereon, discussion of interactions with RSA Data Security, Inc., and consideration of a successor key management RFC's organization. In mail-unrelated discussion, we reviewed activities in the Internet Engineering Task Force's Authentication working group and considered various future topics (e.g., resource access control and policy routing) being discussed in the IAB. John Linn (linn@CCY.BBN.COM) Secretary, Privacy Task Force 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.