Chairman Rob Blokzijl welcomed the participants. Apologies from Thomas Lenggenhager and Ingnacio Martinez. Hans Frese will arrive on the second day of the meeting.
The organization of the meeting was discussed and there will now for the first time be a set of technical sessions. After some restructuring of the agenda was approved.
The meeting decided for future minutes there will be a period of two weeks for commenting after the first release of draft minutes. After this period minutes will be considered finalized. There is the need to make the minutes public available soon after a RIPE meeting.
The previous minutes was approved after one minor change.
Review of previous action list.
Rob Blokzijl:
Antonio Blasco Bonito:
Mats Brunell:
Yves Devillers:
Francis Dupont:
Erik Huizer:
Daniel Karrenberg:
Thomas Lenggenhager:
Matti Rendahl:
Bernhard Stockman:
Marten Terpstra:
Ruediger Volk:
Network Management.
Two main items will be treated, procedures for notification when lines are down and recommendations on IP router management. The idea is to set up routers in such a way that it makes it easier for others to diagnose problems. The working group will try to finalize two draft documents on these subjects presented to the meeting. The working group will be chaired by Daniel Karrenberg.
IP over X.25.
John Seymour volunteered to chair the working group meeting. Currently there are 5 or 6 IP/IXI links going to Amsterdam. More lines will come. Usage of the lines are mainly to
There is a need for recommendations on how to run IP over X.25. There is also a need to have a knowledge on how how much of the IXI traffic that is IP. This should be measured. Two documents are foreseen as output,
Routing on IXI can not be treated as an isolated affair but must be put in conjunction with the global routing.
Network Monitoring and Statistics Gathering.
A proposed agenda for the working group meeting was sent out in advance. There is a need to review the current statistics being gathered. Recent statistical presentations have indicated a need for inclusion of other metrics and additions of new points of measurements. A special problem is the measurement of IP traffic on X.25. There is sometimes a need for a more detailed statistics like level 4 protocol usage. The working group will be chaired by Bernhard Stockman.
RIPE whois database.
Four objects have been suggested for inclusion into the RIPE whois database. An object for Autonomous Systems, an object for Routers, an object for nameservers and an object for network lines. Daniel Karrenberg, chairing this working group, stressed that only objects being proposed in advance will be considered for inclusion.
Domain Name System.
There is a need for specifying placement of European root servers. Flaws have been detected. For example there are top level nameservers with extremely low TTL's. The RFC on DNS gives a wrong indication on the TTL value. Glue records are often unnecessary or erroneous. MX records are sometimes used in an illegal way. There is a need for monitoring the nameservers. Software for monitoring nameservers is available. Two documents are foreseen, a tutorial description of the nameserver configuration and a document regarding the monitoring of nameservers. The working group will be chaired by Francis Dupont.
Bernhard Stockman gave a presentation of the current status of the EEPG work. An interim report on the European network infrastructure was sent out earlier. The line database on which this report is based has changed a lot since that and a revised version is under preparation. Discussions on the possible technologies for a European backbone have started. To investigate the needed services in such a backbone a questionnaire was distributed to the RARE CoA members at the last RARE CoA meeting asking for expected need of network services now and in the near future. Bernhard Stockman expressed the need for input from networking organizations. Daniel Karrenberg proposed that RIPE should give input to the EEPG work.
END OF DAY 1.
DAY 2, entirerly dedicated to working group sessions.
DAY 3 plenary session:
Report to be included.
The IP/X.25 working group:
1. noted that there was currently only one interface between IP/IXI and RIPE, and concluded that there should be more than one.
2. noted that routing between
needed separate and careful study;
3. agreed that in order to plan IP/IXI sensibly, we will wish to have knowledge of the structure of the IXI cloud. The group recommend that this fact be made known to the appropriate authorities, so that any changes made could be decided on with that fact being taken into account.
Planning IP/IXI could realistically cover only the next 9 months or so, because there was inadequate knowledge of the X.25 infrastructure beyond that time.
The group noted that it would be valuable if there was a RIPE backbone with fat pipes joining points in Sweden, Amsterdam, CERN/SWITCH and Italy. It also noted that the principle IXI switches were in Amsterdam and Berne, the latter being convenient to CERN/SWITCH from a networking point of view. Thus, whereas IP not over IXI might "join" RIPE at any of the four points, IP/IXI would "join" RIPE at the point in Amsterdam or at the CERN/SWITCH point. Direct tunneling of IP/IXI should be strongly discouraged, but such routes must in any case NOT be made known to RIPE connected networks.
Invitations need to be sent to the four "points" on the envisaged RIPE backbone. ACTION: RIPE Chairman.
Marten Terpstra and Tony Bates agreed to write documents about IP/X.25 and IP/IXI
They agreed that the first drafts of these documents (a) should have highest priority and should be distributed for comments within three weeks, with a final document being available by mid-May. ACTIONS: Marten Terpstra, Tony Bates.
Marten Terpstra volunteered to give input on routing issues concerning IP/IXI, but time does not allow him to be a full member of this working group.
John Seymour agreed to provide notes of the groups deliberations. ACTION John Seymour.
Report to be included.
The group stated that statistics should be gathered locally within each community but if not possible this could be done centrally. The results, in form of a database with logged data or processed into presentation diagrams, should be made public available. The statistics gathered within the communities must be compatible as there is a need to combine these statistics into global overviews. This stresses the need for a commonly agreed set of tools to be used in the gathering and processing of network statistics. Distributed access to databases with gathered data could be done using SNMP-based tools XLINK currently are developing. The group stated finally that statistics gathered
locally should at least have the same reliability as today centrally gathered statistics.
Points of measurements.
In today complex topology it is necessary to define exactly what we are trying to measure and where. For this reason there there is a need for a model on what and where to measure. Points of interest could be divided into two main categories
Points of type 1 are mainly where US IP traffic enters RIPE. Points of type 2 needs a definition of RIPE communities. As RIPE coordinates European IP networks a community could then be such a network. However countries are the funder of networks and are of course interested in the statistics from the country oriented point of view. If such statistics is required it could then be provided. The group will however for the moment put emphasis on the community oriented view. Bernhard Stockman will write a draft paper giving definition and examples of RIPE communities and adequate points of measurements to be discussed within this group via email. ACTION: Bernhard Stockman.
Restriction on SNMP access
The CERN CISCO is currently using a lot of its CPU-time processing SNMP queries. This is not acceptable and the situation will probably be worse as there is a growing interest for SNMP and public available SNMP software. The group concluded that there is a need for restricting the SNMP access. The group recommends that each community installs their own SNMP access string but makes the result of the statistic gathering public available as described above.
Detailed statistics.
There is sometimes a need to have detailed statistics on a higher layer. For example it could be of interest to see the relative distribution of Telnet, FTP, VMNET, DNS, etc. This could be done using NNstat. To be able to undertake this there is a need of having an NNstat processor installed at each core site. This is currently not the case but could be recommended if there will be a high demand for such statistics. The group will for the moment not undertake any action in this area.
Network monitoring.
A demonstration of a monitoring tool developed by James Barr at NIKHEF was performed at the end of the session day for all participants in this RIPE meeting. This tool seems to give a lot of benefits as compared to today available NMS products. The tool is named "MoNet" and is available via anonymous ftp from nikhefh.nikhef.nl under the directory netmanagement. James Barr would like this tool, being currently a beta-release, to be installed at some sites for testing and enhancement purposes. Questions and ideas should be sent via email to James Barr <[email protected]>.
Report to be included.
Juha Heinanen has suggested that CLNS router information could be included in the RIPE database. Cristina Vistoli objected to that. The meeting concluded that the CLNS group should sort it out amongst themselves. A request to RIPE would be answered positively.
Presentation of the ideas in the document distributed on the first day of this RIPE meeting concerning a technical description of the RIPE NCC. As earlier discussed there is a need to fulfill the items described. This should be done by paid personal and not as now by voluntary efforts which sometimes have their limits.
RARE approved the idea of the NCC and fully supports the idea. RARE EXEC have given the project a high priority and allocated 10 kECU to write down the necessary papers and put them into the right form.
The draft paper currently being discussed is the earlier proposal for the RIPE NCC and the above mentioned technical description. A third paper describing the organizational structure of the RIPE NCC is still missing.
Implementation of the RIPE NCC.
The NCC could be centralized or decentralized. The meeting agreed that it should be centralized. Specific services could be subcontracted. To be able to employ people the NCC must be owned by a legal entity. The NCC could be part of such entity or being outside just using the legal structure of this entity.
Below model was discussed:
There is the need to be able to employ people on a longer period, for example three years. We should look for a friendly organization publically that will support the RIPE NCC. There will probably not be money to do everything we want with the NCC.
In the ideal case we would like to see RARE as the friendly organization, but it could prove necessary for speed's sake to look for another friendly organization at first. Funding is, in all cases, possible from RARE, through RARE or not involving RARE at all. Funding "through RARE" is possible using a RARE accepted method with earmarked money coming from a (sub)set of RARE's contributors.
The conclusion is that we are looking for a friendly organization, preferably RARE. However not to have to wait unnecessary long other organizations will be asked if an answer from RARE is not received within reasonable time.
A document describing above organizational aspects will be written very shortly and distributed to the mailing list for discussion. ACTION Rob Blokzijl.
The time schedule.
The organizational solution and funding should be ready for presentation at the Blois meeting in the middle of May 1991. The NCC should be up and running at July 1. If too much delays in this it is possible that the currently voluntary performed services could not be maintained as is. The exact time scale should be put into the funding seeking paper to ease for funding bodies.
Many eastern countries are requesting IP connectivity to RIPE. The problem is that COCOM regulations still prohibit such connectivity to networks in the USA. In earlier requests from EARN it was stated from the US Department of Commerce that batch oriented traffic was approved while interactive traffic was not allowed.
Currently Poland has requested connectivity to DESY in Hamburg, Poland, Soviet Union, Hungary and Czeckoslovakia have requested connectivity to HEPnet. EARN in Austria has received requests from Czeckoslovakia and Hungary to use the current EARN lines for IP traffic. Italy (Trieste) foresees IP connectivity to the same set of countries. Those countries have also expressed interest to participate in the RIPE initiative.
At the last CCIRN meeting this issues were discussed with US agencies present. The opinion was that these restriction were unnecessary. The regulation may change but for now no East Europe traffic may enter the US part of Internet.
This could however be considered as an US internal affair. They must guarantee for themselves that no East European traffic enters there networks. RIPE members could of course be cooperative and install restrictions on the European end as well.
Decision:
RIPE welcomes IP connectivity to Eastern Europe. RIPE invites representatives Eastern European countries to participate in RIPE activities.
RIPE recommends that if East European networks are connected then US networks and RIPE members operating transatlantic links will be notified so that they take actions they consider apropriate.
It was proposed that a document stating the current situation and problems should be produced. For this reason a small group should be formed consisting of 4 to 5 persons comming from sites heavily involved in international routing. The outcome should be a document that
IBM offers funding for one meeting including routing expertise from the US.
Possible organizations to be represented are HEPnet, CERN, EUnet and NORDUnet. Below people volunteered to participate in this effort:
Willi Porten, being responsible for this action will see to that necessary actions are taken to arrange for a routing group meeting and that specified papers are produced.
Someone from NORDUnet missing. Bernhard Stockman will put out a question to NORDUnet for participation.
Daniel Karrenberg offered to review output from the group.
It was stated that RIPE routers should filter the config files via the RIPE database to give access only to RIPE connected networks.
RIPE strongly discourage intra-European connectivity to go via intercontinental lines and RIPE will coordinate the resolving of these problems. It is also recommended that a warning is sent out. This kind of usage takes away bandwidth paid by other organizations. If the service via US is better than the intra-European connectivity this is an indication of that an upgrade of the European infrastructure is badly needed. In various ways NSF has expressed their concern of European transit traffic back to Europe and have asked that something should be done. European transit traffic not going back to Europe is however allowed.
In the last two day the workplan has been rather loaded. All this work have come out from the revised workplan. The RIPE NCC technical paper will be circulated and discussed on the mailing list.
A proposal for a restructuring of the Task Forces was sent out before this meeting by Bernhard Stockman. The reason as presented in the proposal was that items covered by the Task Forces not always correspond to the current situation and should be reviewed. A group consisting of the RIPE chair and the Task Force chairs will undertake this subject.
IP networking in Belgium. A status report by Pierre Verbaeten, Katholieke Universiteit Leuven.
In most universities and research institutes you will currently find one or more IP islands; mostly UNIX systems interconnected by a LAN. We are now in a phase of interconnecting these islands and providing international IP connectivity.
Currently the 64 Kbps EASInet line from Leuven (Katholieke Universiteit Leuven, K.U.Leuven), to Montpellier provides IP connectivity by running IP on top of SNA. Within one month we also expect to use IP on the line from Leuven (EUnet backbone) to Amsterdam CWI at 19.2 Kbps. Discussions to share a single line are started.
Within Belgium some existing leased lines are currently used to carry IP traffic. This is currently done for lines from Leuven to the Université de Liège and from Leuven to the Universite' Catholique de Louvain. On the longer term a new research network will be used; we hope this one will be operational early 92. Next to IP it will also offer X25 services.
Within Belgium we did already setup a DNS for the top domain .be and servers at 2 universities. As soon as we have IP on the line to Amsterdam, we will make them accessible from outside Belgium.
The conclusion of this meeting is that the meeting structure used was very productive. Subjects involving IXI and RARE were discussed in the absence of represents from these organisations. Invitations sent out for participating in RIPE meetings have always included these organisations.
There was a feeling that the first day plenary session could have been longer to cover most of the main topics.
The RIPE meeting in October in Geneva was earlier decided to be in conjunction with the Telecom'91 conference. However this must be changed as the hotel situation in Geneva at that time already make this impossible.
That meeting is for this reason changed to be at CERN, September 25 to 27, 1991.
ACTION List:
Rob Blokzijl
Tony Bates
Mats Brunell
Daniel Karrenberg
Thomas Lenggenhager
Willi Porten
Matti Rendahl.
John Seymour
Bernhard Stockman
Marten Terpstra