<NIS.NSF.NET> [IMR] IMR88-10.TXT
OCTOBER 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).
BOLT BERANEK AND NEWMAN INC.
----------------------------
September Report
WIDEBAND NETWORK
A new Wideband Butterfly Gateway was installed at the Jet
Propulsion Laboratory in Pasadena, CA on September 21. The gateway
connects JPL ethernet LAN facilities to the Wideband Network via a
T1 access circuit running between JPL and the Wideband node at ISI.
This new network connection has been established to support remote
access to simulations running on a Hypercube located at JPL from
workstations located at USAF/ESD facilities in Lexington, MA. The
connection may also be used to support distributed operation of
simulations on Hypercubes located at the two sites. Completion of
the ESD-JPL Wideband communication path is currently awaiting the
Westine [Page 1]
Internet Monthly Report October 1988
connection of the "ESD-Hypercube" ethernet LAN to the operational
ESD/Mitre Wideband Butterfly Gateway. This connection will be
completed soon.
SATNET
SATNET performance was excellent during August. Statistics
collected by ISI showed an average of 99% uptime for the SIMPs and
98% uptime for the attached gateways. There were only two
problems. On 9/17/88, the Fucino SIMP crashed but was reloaded and
restarted without any difficulty. On 9/28/88, a high error rate
was reported on traffic from Fucino. This was corrected by
increasing the transmit power at Fucino back to normal levels.
Last month, the antenna at the Tanum Earth Station was repointed to
a different satellite thus removing the Tanum SIMP and NTA-RE from
the SATNET. During September, Internet connectivity to NTA-RE was
restored by the installation of a point-to-point link between the
Butterfly gateway at NTA-RE and the Butterfly gateway at RSRE.
October Report
SATNET
This month the SATNET experienced several problems which caused the
overall performance to be less than normal as reported by the tests
from ISI. Goonhilly (93% availability) was down for two days
because of a defective tape drive. Fucino (95% availability)
crashed on 3 occasions and had to be reloaded manually. The cause
is under investigation. The gateway at RSRE was isolated from the
network for 6 days because of a problem on their kilostream link to
London.
Work continues on procuring and installing point to point links
that will be replacing the current SATNET. The new direct link
between RSRE and BBN is now being installed. Testing is expected to
begin in the middle of November.
This month Alex Mckenzie and Claudio Topolcic attended the SATNET
and Infrastructure meeting and the open meeting of the ICB in Pisa.
INTERNET R&D
Our implementation of IP Multicast software for the Butterfly
Gateway is now complete and we have started testing it in our lab.
When it is ready we plan to beta test it some of the Butterfly
Gateways around the Wideband network.
Westine [Page 2]
Internet Monthly Report October 1988
We have installed the hardware and software to support the new
point to point line between BBN and RSRE. We have tested the
hardware and software locally and are waiting for the line to be
activated.
Mike Brescia and Bob Hinden attended the SATNET and Infrastructure
meeting and the open meeting of the ICB in Pisa.
Bob Hinden (Hinden@BBN.COM)
ISI
---
Internet Concepts Project
Greg Finn continues to run simulations of schemes that use Source
Quench within IP to perform end-to-end congestion control. Greg is
also working on a report of the results. Joyce Reynolds and Paul
Mockapetris attended the IFIP WG 6.5 Conference in Irvine, Ca, 10-
12 October, 1988. Joyce Reynolds and Paul Mockapetris also attended
the IETF Task Force meeting in Ann Arbor, Michigan, 17-20 October,
1988.
Three RFCs were published this month.
RFC 1072: Jacobson, V. (LBL), and B. Braden (ISI), "TCP
Extensions for Long-Delay Paths", October 1988.
RFC 1073: Waitzman, D., "Telnet Window Size Option",
BBN STC, October 1988.
RFC 1074: Rekhter, J., "The NSF Backbone SPF based Interior
Gateway Protocol", T.J. Watson Research Center,
IBM, October 1988.
Ann Westine (Westine.ISI.EDU)
Los Nettos
Los Nettos has not yet sent out its first packet but it soon will.
Two lines being installed by GTE were due October 21. However the
lines have not yet been turned up. The last three are due from
PacBell November 4. Datatel CSU/DSU's are delayed in delivery. We
hope for delivery before the last of the lines are installed.
A technical group meeting was held at ISI October 26th. Pacific
Bell gave a presentation on T1 technology and how Los Nettos will
report problems should they occur. We also discussed how to
Westine [Page 3]
Internet Monthly Report October 1988
configure and bring up the network.
During Pacific Bell's presentation they described a new Calif PUC
tariff which will significantly reduce our T1 costs if it becomes
effective in January as planned. The tariff changes could help Los
Nettos grow more quickly as well. Because of the likely tariff
change that could save Los Nettos about $3000 per T1 installation
we are defering orders for phase two lines for Los Nettos until we
can qualify for the new tariff.
Los Nettos is still looking for more participants for phase two of
the network. Our network topology will be developed in phases so
that we can build a well connected, robust, low cost topology. Any
sites wishing to participate in phase 2 should contact Walt Prue
soon.
Discussions are in progress with the University of California
Office of the President to get a donation of some bandwidth between
Los Nettos and the San Diego Supercomputer Center.
Walter Prue (Prue@ISI.EDU)
Multimedia Conferencing Project
This month a teleconference facility was installed at SRI as the
fourth site in the multimedia teleconferencing system. The four
sites (BBN, DARPA, ISI, SRI) are linked by the Wideband Net that
carries both realtime packet voice and video traffic plus datagram
traffic for text and graphics in the shared multimedia workspace.
The new site should provide more opportunities for northern
California folks to participate, and opens the possibility of
four-site teleconferences.
Part of PVP, the butterfly-resident packet video host, was
rewritten in preparation for future variable data rates. Variable
rates will allow better image quality when only two sites
participate.
Work continued on making the multimedia conference control program
(MMCC) more robust. This focused on communication between MMCC and
the Voice Terminal program (VT), on changes to the configuration
file format, and the program's user interface. Additionally,
redesign of autopilot mode was completed. This will make it easier
to control connections remotely and in turn make it easier to get
conferences going for new users at sites away from the experts
(e.g., at SRI).
Westine [Page 4]
Internet Monthly Report October 1988
A demonstration of the multimedia conferencing system was given to
visitors from Concept Communications, suppliers of our video codec.
Steve and Eve attended the TeleCon VIII Conference in Anaheim, CA,
and heard all about the latest in video codecs and teleconferencing
hardware. Steve also attended the Internet Engineering Task Force
meeting in Ann Arbor to participate in the newly-formed ST and
Connection-Oriented IP Working Group.
Steve Casner, Dave Walden, Eve Schooler (casner@ISI.EDU,
djwalden@ISI.EDU, schooler@ISI.EDU)
NSFNET Project
Annette Deschon made improvements and fixed bugs in BFTP. She (1)
updated both BFTP and the BFTPTool to use the latest SRI time
parsing routines written by Ken Harrenstien; (2) incorporated Dave
Curry's fixes to handle suspend and interrupt signals in the BFTP
parser; (3) fixed a bug in which BFTP would give up on a transfer
prematurely in the event that a nameserver was down; and (4) tested
against a number of FTP-server implementations. Dave Mills fixed
Fuzzball FTP bugs she uncovered (we wish ALL FTP implementors were
as responsive to bugs we have found!!) Planned BFTP projects
include: conversion of BFTP and BFTP to Sun OS 4.0, more
documentation, and changes to allow the user to specify that an
alternate directory for BFTP request files.
Annette started work on a version of NNStat for Sun OS 4.0. She
also reviewed a draft of the Hosts Requirements RFC and attended a
one-day Los Nettos meeting held at ISI.
Bob Braden continued work on the Host Requirements RFC, preparing a
new version for discussion at the IETF meeting in Ann Arbor. He
attended the IETF meeting, and chaired a 1.5 day meeting of the
Host Requirements Working Group.
RFC-1072, "TCP Extensions for Long-Delay Paths", authored by Van
Jacobson of LBL and Bob Braden of ISI, was published. This paper
was produced at the behest of the End-to-End task force.
Bob Braden and Annette DeSchon (Braden@ISI.EDU, DeSchon@ISI.EDU)
MIT-LCS
-------
No report received.
Westine [Page 5]
Internet Monthly Report October 1988
MITRE Corporation
-----------------
No report received.
NTA-RE and NDRE
---------------
No report received.
SRI
---
No report received.
UCL
---
We have started work on the design of an expert system to interface
with a simulation system for studying routing and congestion
control problems in Broadband ISDN. Areas of particular interest
are "Mothers Day" synchronised overload, and correlated real-time
traffic.
We have now obtained some primary rate ISDN interfaces, and are
awaiting lines so that we can start experimentation with
interconnecting LANs over ISDN.
John Crowcroft (jon@CS.UCL.AC.UK)
UNIVERSITY OF DELAWARE
----------------------
1. Paul Schragger is using the MIT network simulator to examinine
a priority or type-of-service queueing element in a network
level switch. The model has a high-priority queue and a low-
priority queue and service elements which can serve either or
both queues. The service policy is designed to provide a
ration of high-priority and low-priority service according to
a cost function based on mean service rate and mean delay. He
is currently experimenting with various regimes of arrival
rates and cost functions.
2. Mike Minnich has been reorganizing the campus NTP clock
hierarchy. There are now three stratum-2 servers for the
entire campus, each of which peers with two different
stratum-1 servers elsewhere on the Internet. Each stratum-3
Westine [Page 6]
Internet Monthly Report October 1988
departmental computing cluster or file server peers with all
three stratum-2 servers. Stratum-4 departmental workstation
clients obtain time from their stratum-3 server and may also
peer with one or more other stratum-4 clients. We believe this
scheme is incredibly robust, yet involves only modest peering
overhead.
3. The new telephone-time service provided by the National
Institute of Standards and Technology (nee NBS) was tested
both on IBM PC and Sun workstations. Preliminary results
indicate the NTP stratum-1 (primary) time servers aggree with
the telephone time to within about 10 ms. Mohamed Ellozy of
NIST provided the testing code and is interested in installing
NTP for testing as well. DECWRL has just become the eleventh
NTP primary time server on the Internet. All eleven are
presently chiming truetime. The promised journal article on
NTP is finished except for some diagrams.
4. The 25-year old LORAN-C radionavigation system is being
refurbished with new timing generators capable of
synchronization to UTC with accuracies in the nanosecond
regime. I have suggested the new generators be designed to
phase-modulate the LORAN-C signal with timecode information
similar to that used for the NIST WWV/WWVB/GOES services. This
would provide many more transmitters at much higher powers
than the NIST services and possibly result in cheaper timecode
receivers with less demanding antenna instalations.
5. Jeff Simpson is still working on his logical-list processor
for policy-based routing and a paper explaining how it all
works. Chuck Cranor has completed the design for a 3x3 VLSI
digital crossbar switch as a class project. He is working with
Paul Schragger on possible extensions to his design and
possible application to a high-speed reservation-based packet
network. Mike Davis is busy scarfing up the NSFNET Phase-I
Backbone performance data lying around the Interet for use in
a class project. Dave Mills worked out the entropy and some
hashing properties of Internet address assignments and played
the issues for senior gurus. Dave also participated with
Annette DeSchon of ISI and Joel Gartland of FTP Software in
tests involving three-way FTP transfers with generally joyful
results.
Dave Mills (Mills@UDEL.EDU)
Westine [Page 7]
Internet Monthly Report October 1988
NSF NETWORKING
--------------
NSF NETWORKING
UCAR/BOLT BERANEK AND NEWMAN INC., NNSC
Craig Partridge attended the Internet Engineering Task Force
meeting where he chaired a Management Information Base working
group session and participated in the Internics meeting.
Karen Roubicek attended the ACM SIGUCCS meeting on User
Services and ran a Birds of a Feather session on NSFNET.
The fifth issue of the NSF Network News was published and
distributed at the end of October. For additional copies,
please send a message to nnsc@nnsc.nsf.net.
by Karen Roubicek (roubicek@nnsc.nsf.net)
NSFNET BACKBONE (MERIT)
October again shows a steady increase in traffic on the NSFNET
backbone.The number of networks with primary connections to
NSFNET through the mid-level networks has gone from 292 at the
end of September to 305 at present.
__________________________________________________
Packets in Packets out
September 314,675,718 304,171,588
October 443,389,133 356,779,596
% increase 29.0% 14.7%
___________________________________________________
Packet counts are taken at the token ring interface to the E-
PSP in each Nodal Switching Subsystem (NSS) via SGMP. The
hourly counts are collected and stored in a database on the
Information Services host machine. Information from this
database is planned for public availability in November.
Other databases planned for public release within the next few
weeks are now in the final stages of staff testing. These
include information on site contacts, routing data, and
general documentation.
Major planning efforts are underway to develop a comprehensive
schedule for upgrading the Nodal Switching Subsystems and
Network Management Subsystems in the next year. These are
Westine [Page 8]
Internet Monthly Report October 1988
aimed at implementing the goals outlined for Phase 1A in
Merit's proposal to NSF.
Merit's Susan Hares organized an IETF/FARNET working group to
discuss how the NSFNET backbone, regional networks, and campus
networks could cooperate to solve user problems expediently.
The first meeting of this group, the IETF Joint Monitoring
Access for Adjacent Networks (NSFNET-JoMAAN), focusing on the
NSFNET community was held October 17, 1988 during the IETF
meeting in Ann Arbor, MI. Susan chairs the group and twelve of
the thirteen regional sites sent representatives to the
meeting. Additional representatives from NSF, Arpanet, DDN and
NSN participated. NSFNET site representatives also gathered at
a luncheon during the IETF meeting to discuss areas of common
concern with the Merit/NSFNET staff.
A number of Merit/NSFNET staff members participated in Educom
88 in Washington, D.C.during October. A fully-functional NSS
was on exhibit and used to provide Internet connections to
conference participants. A group of Internet addresses have
been permanently assigned for use by this traveling node
during future conferences and trade shows.
by Ellen Hoffman (Ellen_Hoffman@um.cc.umich.edu)
NSFNET BACKBONE SITES & MID-LEVEL NETWORK SITES
BARRNET
No report received.
CERFNET
No report received.
CORNELL UNIVERSITY THEORY CENTER
No report received.
UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN/NCSANET
No report received.
JOHN VON NEUMANN NATIONAL SUPERCOMPUTER CENTER
No report received.
Westine [Page 9]
Internet Monthly Report October 1988
MERIT/UMNET
Merit has continued to refine the Merit INP over IP
implementation.
Submitted by Merit/NSFNET Information Services (NSFNET-
info@merit.edu)
MIDNET
No report received.
MRNET
There are several new officers of MRnet. The MRnet officers are:
o Mahlon Stacy, Chair
mcs@mayo.edu
o Dan McCreary, Vice-Chair
dan@gonzo.eta.com
o Carl Henry, Treasurer (replaces Peter Pyclik)
chenry@carleton.edu
o Tim Salo, Secretary (replaces Ken Carlson)
tjs@uc.msc.umn.edu
MRnet is preparing for expansion. The MRnet network is in
place; the next objective is to offer network services to more
organizations. An MRnet brochure has been printed to assist
in attracting new members. MRnet is also raising funds to pay
for the link to the NSFnet during 1989 and thereby provide a
reliable Internet connection.
With the recent attachment of St. Olaf College to MRNet, we
are now announcing the following member networks:
128.101.0.0 UMN-NET U of Minnesota, Minneapolis
U of Minnesota, Duluth
Minnesota Supercomputer Center
192.35.86.0 UMN-MORRIS-NET U of Minnesota, Morris
129.28.0.0 ETA-LAN ETA Systems
129.176.0.0 MAYO Mayo Foundation
129.191.0.0 NSCO Network Systems Corp.
129.205.0.0 CDCNET Control Data Corp.
130.71.0.0 STOLAF St. Olaf College
192.12.250.0 MRNET Minnesota Regional Net hub
Westine [Page 10]
Internet Monthly Report October 1988
as well as forwarding mail for Cray Research (CRAY.COM).
by Tim Salo (tjs@uc.msc.umn.edu)
Stuart Levy (slevy@uc.msc.umn.edu)
NATIONAL CENTER FOR ATMOSPHERIC RESEARCH AND UNIVERSITY
SATELLITE NETWORK PROJECT
USAN was down for eight hours in October for Vitalink to test
its satellite transponders.
Sometime in November, Los Alamos National Lab (AS #68) will
EGP peer with the NCAR NSS and directly connect to the NSFNET,
nets 192.5.16, 192.16.16, and 128.65 via a 56Kbit link and
cisco boxes.
Still unexplained is the additional 100 msec delay for round-
trip times over the satellite links that occured after the
upgrade to Comstream modems earlier this year. The modems were
installed at all sites to correct a drop-out problem. Current
RTT's are now in the neighborhood of 640 msecs.
by Don Morris (morris@windom.ucar.edu)
NORTHWESTNET
The NorthWestNet annual meeting was held October 10-12 in
Beaverton, OR.
Two new commercial members, Battelle (DOE Richland WA) and
Intel Corp. (Beaverton OR) have joined NorthWestNet.
by JQ Johnson (jqj@oregon.uoregon.edu)
NYSERNET
No report received.
OARNET
No report received.
PITTSBURGH SUPERCOMPUTING CENTER
PSCnet is carring traffic for several new campuses, including
Univ. of Toledo (192.42.112), a new Penn State network
(130.204), and University of Akron (130.101). There has been
no significant down time for our connectivity to the NSFnet.
Westine [Page 11]
Internet Monthly Report October 1988
The T1 link to SURAnet is under heavy use: carrying almost 1
Million packets per day. Since this line parallels the
NSFnet, it is considered redundant and its continued support
is in question.
We are preparing the PSC internal network for the installation
of the new Cray Y-MP in mid December. We will be providing
connectivity directly to the Cray via both HYPERchannel and an
FEI-3. No network related service outages are expected either
for Cray X-MP access of for PSCnet through traffic.
by Matt Mathis (mathis@fornax.ece.cmu.edu)
SAN DIEGO SUPERCOMPUTER CENTER
No report received.
SESQUINET
On October 17th, Guy Almes chaired the second meeting of the
IETF Working Group on Interconnectivity. We focussed on
getting feedback from IETF members at an open session, then
working to revise EGP-3 during a closed session. We are
specifically working to liberalize the "core" model of EGP-2
to better meet the demands of the current NSFnet Model with
multiple national backbones, mid-levels, and campus networks.
The complete initially proposed SesquiNet configuration has
been operational for a year now. The following campus
networks are being served, and are advertised via EGP to
NSFnet and (currently via UIUC) to the Arpanet core:
Baylor College of Medicine 128.249
BCM-Technologies 192.31.88
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
Westine [Page 12]
Internet Monthly Report October 1988
In addition we are advertising to NSFnet the following
networks in cooperation with the University of Texas:
UT-Austin 128.83
UT-HSC-Houston 129.106
UT-Arlington 129.107
UT-ElPaso 129.108
UT-MB-Galveston 129.109
UT-Dallas 129.110
UT-HSC-SanAntonio 129.111
UT-HSC-Dallas 129.112
UT-PermianBasin 129.113
UT-CCSPRD 129.114
UT-CHPC-Hyperchannel 129.116
Texas Tech University 129.118
University of North Texas 129.120
UT-SanAntonio 192.6.201
THEnet 192.16.72
and UT-Austin-TestNetwork 192.16.73
The new NSFnet backbone node at Rice University became
operational during the last week of June, and has proved quite
reliable. FTPs of 96kb/s across the new NSFnet are typical.
The triangle connecting UT-Austin, Texas A&M, and Rice
University is now up and operational.
by Guy Almes (almes@rice.edu)
SURANET
No report received.
Westine [Page 13]
Internet Monthly Report October 1988
WESTNET
1. Testing of SNMP was begun at the University of Colorado
at Boulder. After experience is gained, we plan to run
it at selected locations within Westnet.
2. The Westnet Steering Committee acted to form two
subcommittees. The first is chaired by Kelly McDonald of
Brigham Young University, and will assess Bitnet issues,
including a feasibility assessment of routing Bitnet
traffic over digital circuits within Westnet. The second
is chaired by Bob Leach of the University of Arizona, and
will attempt to establish appropriate divisions of
responsibility of NIC/NOC activity among the campuses and
central Westnet staff.
3. The circuit between New Mexico Technet (NMT) and NCAR was
subject to an unacceptably high error rate (10%). A
variety of strategies were implemented to alleviate this
problem, Some of which will be enumerated here. (1) Cisco
reported that the cables between serial ports and
CSU/DSU's should not exceed 5' in length. We had a number
of cables which exceeded this length (some by half an
order of magnitude). This was corrected, and seemed to
decrease the error rate by on the order of one percent.
(2) Some of the CPU ROM's, ver. 6.1(626), in the ciscos
were out of date. The CPU ROM's were upgraded to the most
recent (ver. 2.9), and this reduced the error rate by a
similar amount. (3) Finally, new ROM's (ver. 2.9) were
installed in the serial boards. This resulted in much
improved performance (error rate of less than one
percent). New ROMs for the serial boards have been
ordered for all Westnet nodes.
4. This has resulted in some interesting tests being done by
Carol Ward and David Wood. They have measured the error
rate on one serial port when doing a local loopback while
the other port was: disconnected, connected, and routing
was turned off. They observed some inexplicable
interdependence of the error rate, depending on what is
happening with the other serial port. We are in contact
with cisco, and will keep the pressure on.
by Patrick J Burns (pburns@super.org)
Westine [Page 14]
Internet Monthly Report October 1988
TASK FORCE REPORTS
------------------
APPLICATIONS -- USER INTERFACE
The task force met September 12-13 at ANSA Headquarters in
Cambridge, England. We took this opportunity to tap the
Cambridge computer science community, with observers from
EuroPARC, Olivetti Research Ltd., and the Cambridge University
Computer Laboratory (in addition to our own long-standing
member from ANSA, Joe Sventek). With a resulting attendance
of 18, 9 of whom were observers, this meeting had a distinct
"workshop" flavor.
Close to half the meeting was spent discussing user interface
architecture. Several people entered the meeting believing
their architectures were radically at odds with those proposed
by other parties, but left the meeting believing the
architectures were in fact quite similar. For example, there
was general agreement that we need something considerably more
powerful than user interface toolkits such as Xt and Andrew.
Such toolkits are particularly deficient with respect to
support for multi-user interactions---e.g. two participants in
a real-time conference cannot see two different views of the
data. Indeed, this same deficiency also prevents users of
existing toolkits from altering, for example, the layout of
related windows on the screen without modifying the associated
application(s); yet, it should be possible, for example, for
one user to have all his scrollbars or button menus on the
right whereas another user has them on the left. While
different people used different terminology to address this
problem, the common approach was to introduce another level of
abstraction "above" that of Xt's widgets or Andrew's views.
In my own case, for example, this leads to a distinction
between "application interface objects" and "presentation
objects", with the latter being roughly equivalent to Xt
widgets and the former providing complete isolation between
data and view (necessary, for example, in order for two
different participants in a conference to see two different
views of the data simultaneously).
Ultimately, there was general agreement with my "reminder" of
our previously agreed-upon distinctions between reference
models, architectures, and implementations---where there may
be many different implementations of a given architecture and
many architectures that conform to a given reference mode---
and with the related claim that more work was needed on all
three fronts. To that end, I have instituted a smaller
Westine [Page 15]
Internet Monthly Report October 1988
working group to develop a strawman (or strawmen), which
effort can be regarded as the second major joint effort of the
task force to date (the first being voice servers).
Other highlights of the meeting included:
- There were several nice contributions with respect to
floor control (in the context of real-time
teleconferencing). For example, Earl Craighill (SRI
International) presented a layered model of real-time
multi-user interactions wherein the floors for each
interaction mode ("read only" discussions, foreground
updates, and background updates) could be dealt with
separately. He then discussed four specific floor control
algorithms for use in various contexts, the most
interesting being their support for interjections---
specifically, permitting any party to speak (and be heard)
for some fixed quantum of time, even if the person with the
floor does not give it up. While this algorithm cuts
interjections off (if the person with the floor does not
stop speaking), it is more user friendly than the usual
"only one person may speak at a time" algorithm.
Subsequently, Lester Ludwig (Bellcore) discussed some
"audio windowing" techniques that permit computer-based
conferencing to come even closer to emulating face-to-face
meetings---by permitting all parties to speak at once, but,
via special audio effects, to permit one party to be heard
"better" or "more easily".
- High-quality audio/video is all the rage. Nine of the
organizations represented at the meeting (Olivetti Research
Center, Olivetti Research Ltd., ISI, Xerox PARC, Xerox
EuroPARC, Bellcore, Sun, HP Labs, and Cambridge
University) have developed or are developing extensive
audio/video testbeds, all computer-controllable, but using
a variety of analog and digital networks. Two major items
on most of these groups' wish lists are networks capable of
adequately supporting (digital) audio and video, and a
cheaper alternative to the Parallax graphics+video board
(which is the only currently available product capable of
providing live video in a window). Not surprisingly,
several of these same organizations are working to satisfy
both these wishes.
- Also with respect to video, Phil Gust (HP Labs) has
started an effort to extend/adapt Olivetti's VOX audio
server to support video.
Westine [Page 16]
Internet Monthly Report October 1988
- Based on Bill Buxton's overview and a subsequent visit by
the task force to their offices, Xerox EuroPARC is
definitely a place to watch---in the domains of human-
computer interaction and computer-supported cooperative
work.
- Ditto for Olivetti Research Ltd. (trying to be objective
here!)--- in the domain of networking.
Keith Lantz (LANTZ@ORC.OLIVETTI.COM)
AUTONOMOUS NETWORKS
The ANTF will meet jointly with Barry Leiner's November
Workshop. A second meeting is planned for February in
conjunction with the Privacy Task Force.
Deborah Estrin (Estrin@OBERON.USC.EDU)
END-TO-END SERVICES
The End-to-End Task force met at Stanford University on August
16, 1988, one day prior to SIGCOMM'88. The group will meet
again on November 3-4 at MIT LCS.
Recent Task Force efforts have concerned the following areas:
A. IP Multicasting
1. Plan: make useable implementations of IP multicasting
widely available.
(a) Host implementation for 4.3+BSD.
A prototype implementation has been done, and we
are trying to get it meshed into the BSD sofware
releases.
(b) Multicast router implementation for RIP-based
gateway. A prototype implementation was done at
BBN, although it needs more work and is awaiting
further funding. An RFC describing the protocol
has been submitted and should be published soon.
(c) Multicast router implementation for SPF-based
gateway. An implementation of multicast routing
for butterfly gateways is in progress at BBN.
Westine [Page 17]
Internet Monthly Report October 1988
2. Plan: nurture applications of IP multicasting.
Most of the current interest in using IP multicasting
is to invoke local-network multicasting; for example,
the OSPFIGP group is considering its use for
multicasting routing updates. Local network
applications are important and immediately useful,
although we believe that trans-gateway multicasting
will lead to more exciting new applications.
B. Transaction Transport Protocols
1. Plan: Make VMTP readily available for experimentation in
a variety of contexts.
A release of VMTP for a 4.3+BSD or Sun OS 4.0 system
is scheduled for November 1988.
C. Performance Issues
1. Plan: Develop specifications for experimental TCP
performance extensions.
This was accomplished with the publication of
RFC-1072, "TCP Extensions for Long Delay Paths".
2. Broad research question: can adaptive rate-based flow
control be made to work stably and fairly, as
window-based flow control now does?
3. Broad research questions: the pros and cons of onboard
protocol processing vs. outboard processing, and the
necessity of special "high performance" transport
protocols.
4. Broad research question: The organization of the
transport layer and its interface to the application
layer. For example, where should marshalling/
demarshalling, checksumming, and data copying be
performed for maximum performance?
5. Other areas of concern:
* Applying transport performance ideas to other
protocols, especially UDP-based applications.
* Developing useful tools for measuring and
simulating transport protocol performance.
Westine [Page 18]
Internet Monthly Report October 1988
* Performance (as well as "power") of known data
structuring protocols (ASN.1, Courier, etc.).
Bob Braden (Braden@ISI.EDU)
INTERNET ARCHITECTURE
The next meeting of the INARC Task Force will be held in
conjunction with the IAB meeting in Santa Clara, CA, on 10-11
January 1989. The workshop will include invited presentations
by research contributors from throughout the Internet
community and especially the IAB and its task forces.
Important areas of research interest include policy-based
technologies, advanced routing architectures, high-speed
networks and interfaces, network management, congestion
avoidance/control and advanced transport protocols.
Present plans for the first day are for the IAB task-force
chairs to present in-depth summaries of past activity and
anticipated future work. This will include an assessment of
other ongoing work in the area, with special emphasis on
issues affecting the growth in size, scope and
interoperability of the Internet. On the second day volunteers
are solicited to present concise papers of 20-30 minutes in an
area of specialization. Appropriate papers may be selected for
publication in the ACM Computer Communication Review.
Volunteers do not have to be members of the IAB, its task
forces or their dependents. Prospective attendees do not have
to volunteer a paper, but they must expect to be harassed and
caught up in lively discussions. Please send a note expressing
your interest and/or paper topic to mills@udel.edu.
Dave Mills (Mills@UDEL.EDU)
INTERNET ENGINEERING
1) The IETF met on October 17-19 at the University of Michigan
in Ann Arbor. The meeting was hosted by Hans-Werner Braun and
Elise Gerich of Merit. The agenda, as executed at the
meeting, is given below. Proceedings for the June USNA
meeting were distributed at Ann Arbor and will be available
from the NIC.
Westine [Page 19]
Internet Monthly Report October 1988
2) The final agenda of the Oct 17-19 meeting:
MONDAY, OCTOBER 17
9:00 am Opening Plenary, Introductions and local arrangements
9:30 am Working Group Morning Sessions
o Host Requirements, Members Only (Braden, ISI)
o ST and Connection-Oriented IP (Topolcic, BBN)
o CMIP-Over-TCP Net Management (Lee LaBarre, MITRE)
o Interconnectivity and EGP3 (Almes, Rice)
o Open SPF IGP (Petry, UMD and Moy, Proteon)
1:30 pm Working Group Afternoon Sessions
o Host Requirements, Open (Braden, ISI)
o ST and Connection-Oriented IP (Topolcic, BBN)
o CMIP-Over-TCP Net Management (Lee LaBarre, MITRE)
o Interconnectivity and EGP3 (Almes, Rice)
o Management Information Base (Partridge, BBN)
5:00 pm Recess
7:30 pm o Working Group for Joint Monitoring Access for
Adjacent Networks focusing on the NSFNET
Community (Hares, Merit)
TUESDAY, OCTOBER 18
9:00 am Opening Plenary
9:15 am Morning Working Group Sessions
o Host Requirements, Members Only (Braden, ISI)
o TELNET Linemode (Dave Borman, Cray)
o Authentication (Schiller, MIT)
o Performance and Congestion Control (Mankin, MITRE)
o Point-Point Protocol (Perkins, Hobby, Prindeville)
o PDN Routing (Rokitansky, FernUni Hagen)
1:00 pm Opening Plenary Statement (Gross, MITRE)
1:15 pm Network Status Reports
o Merit NSFnet Report (Braun, UMich)
o IBM NSFnet Report (Drescher, IBM)
o Arpanet/DDN Report (Lepp, BBN)
o Internet Report (Brescia, BBN)
o Interop 88 Network Report or `How to build
a complex internet in 2 days' (Almquist)
3:45 pm Network Performance Presentations
o Packets Over A Different Kind Of Ether,
including Amateur Packet Radio Demonstration
(Karn, Bellcore)
o Keeping The Usual Ether Filled Up With High
Performance TCP (Jacobson, LBL)
5:00 pm Recess
7:00 pm NSFNET NOC Tour
Westine [Page 20]
Internet Monthly Report October 1988
WEDNESDAY, OCTOBER 19
9:00 am Congestion Control Observations Using NETMON
(Mankin, MITRE)
9:30 am Working Group Reports and Group Discussion
o Authentication (Schiller, MIT)
o CMIP-over-TCP (CMOT) (LaBarre, MITRE)
o Interconnectivity (Brim, Cornell)
o Host Requirements (Braden, ISI)
o Internet MIB (Partridge, BBN)
o Joint NSFNET/Regional Monitoring (Hares, Merit)
o Open SPF-based IGP (Petry, UMD)
o Open Systems Routing (Lepp, BBN)
o PDN Routing (Rokitansky, FernUni Hagen)
o Performance and CC (Mankin, MITRE)
1:00 pm Working Group Reports and Group Discussion (cont)
o Pt-Pt Protocol (Perkins, CMU)
o ST and CO-IP (Topolcic, BBN)
o TELNET Linemode (Borman, Cray)
1:45 pm What is Usenet?, What Is NNTP? (Spafford, Purdue)
2:30 pm The NIC Domain Chart (Lottor, SRI-NIC)
2:45 pm On Some T1 Satellite Link Performance (Lekashman, Ames)
3:15 pm Concluding Plenary Remarks
3:30 pm Adjourn (Rush to Airport)
3) The following chart gives a synopsis of IETF Working Group
progress, where `NA' means `Not Applicable' and `-' means `No'.
The three groups marked with `*' are newly formed since the June
USNA meeting.
This type of status report will be updated and reported in future
Internet Monthly Reports. For additional information on IETF
Working Group activities, send a note to ietf-
request@venera.isi.edu.
Active Charter RFC or Met at USNA Met at
Working Groups Submitted? IDEA? USNA? Report? Ann Arbor?
----------------------------------------------------------------------
Authentication Yes Yes Yes Yes Yes
CMIP-over-TCP (CMOT) Yes Yes - - Yes
Interconnectivity* Yes NA NA Yes
Host Requirements Yes - Yes Yes Yes
Internet MIB Yes Yes Yes - Yes
Joint NSFNET Monitoring*Yes NA NA Yes
Open SPF-based IGP Yes Yes Yes Yes Yes
Open INOC Yes - Yes Yes -
Open Systems Routing Yes Yes Yes - -
PDN Routing Group Yes Yes Yes Yes
Westine [Page 21]
Internet Monthly Report October 1988
Performance and CC - - Yes Yes Yes
Pt-Pt Protocol Yes Yes Yes - Yes
ST and CO-ip* Yes Yes NA NA Yes
TELNET Linemode Yes Yes Yes Yes Yes
------------------------------
Groups with completed missions
------------------------------
Domain - Yes Yes - NA
EGP3 Yes Yes - - NA
OSI Technical Issues Yes Yes - - NA
Short Term Routing Yes Yes Yes Yes NA
SNMP Extensions - Yes Yes - NA
Phill Gross (gross@gateway.mitre.org)
INTERNET MANAGEMENT
No report received.
PRIVACY
Minutes from the September Privacy Task Force meeting were
distributed to the task force and to the privacy-interest
mailing list. Some work was performed on updates to the key
management RFC based on September meeting discussion and on
inputs from RSA Data Security, Inc. The provisional 6
December RFC review session remains provisional at this time,
pending the status of the revision effort. Plans for the next
full-scale task force meeting, to be held February 14-16, 1989
at ISI (partly in joint session with the Autonomous Networks
Task Force) were stabilized.
John Linn (Linn@CCY.BBN.COM)
ROBUSTNESS AND SURVIVABILITY
No report received.
SCIENTIFIC COMPUTING
No report received.
Westine [Page 22]
Internet Monthly Report October 1988
DSAB - Distributed Systems Architecture Board
---------------------------------------------
The Distributed Systems Architecture Board has several active task
forces. This report contains a summary of the efforts of the past
few months of two of them.
A Report from the User Interface Task Force Report: Keith Lantz,
Olivetti Research Center, Menlo Park, CA
The task force met this summer at Olivetti Research Center in Menlo
Park. The major topics of discussion were group work, user
interface architecture, video, and real-time requirements of voice
and video. The resulting highlights were
- Andy Schulert of On Technology (and formerly principal architect
of Open Dialogue at Apollo) added his support to the argument that,
in the world of "network computing", "user interface toolkits"
would be better placed on the same machine with the window server.
This effectively means encapsulating them in a separate server (as
is being doing at Olivetti, for example) or in the window server
itself (as several groups are doing with NeWS).
- Greg Foster of Xerox PARC noted that the Colab group is moving
away from "electronic meeting rooms" per se to the more general
concept of "shared workspaces". The idea is to provide support for
the sorts of serendipitous meetings that occur throughout the day-
--in the lounge, by the coffee pot, and in people's offices---
rather than require people to pick up all their notes and move to a
dedicated room to have a meeting.
As for video, T1- (also CDI-) quality implies 200 Kbytes/sec
throughput. As with voice, this presents no problem bandwidth-
wise, but rather with regard to response (to request to store or
retrieve individual frames). However, true broadcast TV requires
roughly 60 Mb and studio quality requires 140 Mb bandwidth, either
of which present significant problems for UNIX.
Steve Casner reported on the Video Working Group Media Lab. First,
since video is the newest kid on the block, a taxonomy of uses
(and advantages over other media) has yet to emerge. Second, there
appeared to be a consensus that doing video "right" is all the
harder because TV has made people accustomed to relatively high
quality; that is, expectations are so high. Third, there was a
feeling that researchers may be placing too much emphasis on either
achieving, at low speeds, quality currently achievable only at T1
rates, or achieving much better quality without regard to bandwidth
requirements, rather than giving users the ability to select
Westine [Page 23]
Internet Monthly Report October 1988
between quality and cost. Fourth, there is a clear need for
authoring systems.
- As for voice server work, projects continue at the Sun New Media
Workstation Group to create a full-function PC- based server using
the TI speech board and at Olivetti Research Center on a software
"audio server" that does for audio what window servers do for
graphics (and then some --- e.g. computer-controlled mixing and
routing). The design specification on the latter will be
available shortly.
A Report from the Naming Task Force: Larry Peterson, University of
Arizona
Karen Sollins, Mic Bowman (one of my Ph.D. students), and I have
have been working on the RFC describing UNP (Universal Naming
Protocol). The going has not been as fast as we'd like, but we have
been able to hammer out most of the details. The plan is to have a
draft specification and an example implementation done sometime
this fall.
Members of the naming task force, as well as several other
interested individuals, have been studying X.500 and comparing it
with UNP. The consensus is that UNP has a more flexible data set
(especially its type system) and supports a richer functionality
(i.e., supports client-defined resolution functions). It also seems
to be the case that UNP is more fully specified (or at least will
be once the specification is done); it's not at all clear that two
independent implementations of X.500 will be able to talk to each
other.
Charlotte Tubis (Tubis@Purdue.Edu)