RIPE 3 took place at NIKHEF on 22-23 January 1990.
CERN, IBM and BelWu apologized for not being able to attend the meeting but reconfirmed their support for the RIPE initiative.
A draft of the minutes was distributed and was accepted after a number of corrections were made.
Since the last RIPE meeting two important events had taken place, the CCIRN meeting that gave the US members a better perspective on the European situation and the setting up of a learned committee by RARE to study IP networking in Europe.
The CCIRN meeting took place at CERN on the 6/12/89. The main topic discussed was IBM's EASInet initiative. EASInet supports IP and proposed that RIPE be represented at the meeting. A second meeting was held later the same day in which Rob Blokzijl gave a presentation on RIPE's objectives.
The "wise men" committee was set up by the RARE executive. The intention of this group was to study IP in Europe and produce appropriate recommendations to be acted upon by RARE.
All but one of the recommendations were seen as important for RIPE with the third, regarding RIPE as the appropriate body for coordinating IP activities in Europe, as the most important.
RIPE fully supports the second recommendation, that an all encompassing European networking strategy should be developed.
The feeling, however, was that it was inappropriate in the context given. It was stated that such a strategy should not exclude any existing networks or technology but that proprietary systems should be depreciated.
The COA will be asked to give RIPE liason status so that RIPE has a definite voice. It was also decided that RARE would be asked to recommend that RIPE be represented on the CCIRN, independently of RARE itself. It was thought that the COSINE policy group would be the best people to approach.
RIPE members agreed that the current membership model of RIPE should be kept. Organisations offering IP services should be members of RIPE. Service providers, such as PTT's or IXI, do not necessarily have to play a role in RIPE.
Carl-Herbert Rokitansky gave a presentation on the IETF-PDN routing working group. The idea of mapping IP addresses to X.121 provoked some discussion on the practicalities of implementing such a scheme. Details of the scheme were distributed at the meeting.
5.1. UK
Paul Bryant reported that there was some interest in IP networking in the UK, although the JNT has no current policies in this area. UK Space Scientists, in particular, would like international IP connectivity. In addition, he estimated that a very high percentage of UK sites supported IP locally.
A number of leased lines from Rutherford to the rest of the continentare planned or exist already. These would carry IP as soon as there was a demand for it.
Proper IP addressing has been used within the UK, so that connectionto Europe ought to be technically straight forward.
Apparently JNT had no intention of letting the fat pipe from the US tothe UK be used for continental traffic. It was decided that the US side would be contacted to clarify the situation.
5.2. Germany
A 64K satellite link from WIN to NSFnet would be available by summer1990, although the division of technical, administration and costing aspects had not yet been finalised.
The line, supporting both IP and DECnet, would be between the Lawrence Livermore Laboratory and the Max Plank Institute at Munich.
5.3. European IP Maps
CERN and Thomas Lenggenhager had produced preliminary maps of European IP networking. The general feeling towards their efforts was very positive.
A number of corrections were made and there was some discussion on how to improve their format and information content. There was some deliberation on whether or not to include internal connections, internal connectivity or just international gateways. It was felt that some effort should be spent searching for tools to produce the maps automatically.
Countries missing from the maps were Spain, Belgium, Portugal, Greece, Austria, Eire and Luxembourg.
Each group discussed its progress with all participants present.
6.1. Task Force 1: Connectivity and Routing
Reports were given on the progress of each task and was followed by a discussion on charging.
1-1: Inventory of International IP Connectivity. Bernhard Stockman reported making a survey of IP networks in Europe by polling SNMP agents. The survey only included those parts that were visible and so excluded most of the United Kingdom and parts of the Netherlands. France seemed to be quite complicated.
In order to gain more information about individual networks a questionaire had been sent out, its format, however would need to be improved.
A need to enumerate the topology of particular subnets in the database was expressed. At that time three distinct types had been identified, these were: backbones, centralised clusters and non-centralised point to point networks.
Clarification by more rigorous use of class B networks was thought infeasible because renumbering existing class C networks would be too much work, although some present thought it could be done.
Thomas Lenggenhager expressed the view that this was a separate issue and that the current intention ought to concentrate on describing the existing layout and leave improvements until later.
1-2: Inventory of unconnected national IP infrastructures. The infrastructures of some countries were completely unknown. It was thought important that at a minimum the names of national contacts ought to be known.
Talks were planned between NIC, MERIT and RIPE to coordinate the exchange of database information.
1-3: Inter continental interconnectivity. A definitive document would be presented at the next meeting.
1-4: International routing scheme design. A definitive document would be presented at the next meeting.
1-5: Monitoring of routing coordination. Some work done.
A lengthy discussion followed a suggestion that some comments be madeon how charging would be implemented within RIPE. Because charging was an extremely difficult problem, a fair scheme was thought to be fairly impos-sible to implement, requiring extensive route filtering.
The internet charging model was thought to be impractical in Europe because no centrally funded backbone existed as it did in the United States. A number of suggestions were made on how a European Internet ought to be designed. There was a strong feeling, however, that maximal connectivity ought to be achieved first. Any divergence from this plan would likely result in too many complications and prove counter productive. The goal would be to achieve total connectivity on the top domains and then improvements could be developed from this base. One solution would be to guarantee some links as openly available which would define a European backbone and hence the US Internet model would be appropriate.
This would require a dialogue between task forces one and four. It was also felt that an effective IP strategy would have to be planned with regard to the rest of European networking. Thus, although a better design was attractive it would be difficult to optimize because changes infavour of IP users may adversely affect traffic using other protocols on the same line.
It was observed that a sizable part of RIPE's function was as a clearing house to balance costs. RIPE's role would be to suggest changes, not enforce them, because primarily RIPE did not intend to provide any services itself. In conclusion, it was stated that financing should be done by bilateral agreements, which in effect had nothing to do with RIPE directly.
RobBlokzijl stated that at present, a policy of no restrictions would be best, and participants should verify this with their own organisations. He also made clear that RIPE was not in any sense looking for an optimal system but was merely an attempt to make something from chaos.
Progress reports on each task were presented, followed by a number of suggestions on improving the database format; a description of how the whois database was to be updated and an explanation of the intended use of the RIPE distribution lists. Finally it was reported that Phil Gross, the head of IETF(Internet Engineering Task Force), had invited Europeans with IP networks to all their meetings. There was a feeling that such contacts with the US Internet would be beneficial and should be encouraged.
2-1: 'whois' database. Daniel Karrenberg reported that the whois software was up and running and that the sources were available by anonymous ftp from mcsun.EU.net. A script was included for building tables for Ciscos from the database information and a query by mail feature would be implemented if anyone needed it. It was also thought that the software should be made available from a number of other machines too.
2-2: Operational contacts infrastructure. This work was thought very important and it was promised that it would be done by the following week.
2-3: Security notification procedure. Should have been done by the following week.
2-4: Common operations procedure. Some work had been done but no formal results had yet been attained.
2-5: Gathering of operational statistics. Bernhard Stockman presented some of the results obtained using a suite of tools for analysing data obtained using the SNMP protocol. The utility of this software was illustrated quite sharply by the histograms presented. It was reported that Marten Terpstra was counting hosts connected to RIPE and estimates would be sent to the ripe-org list. The estimate at that time stood at 12800.
2-6: European Network Information Centre. No report would be given until March.
2-7: Networking tool box. The distribution and availability of software tools had been carried out quite successfully.
2-8: Centres of expertise. No report given. The database topics discussed are summarised below.
In regards to task 1-1, there was the question of how to represent different network topologies in the database.
There was a problem with distributing updates, either by email or ftp. The existing procedure used mail (to [email protected]), with networks sorted by IP address and people by surname. The possibility of replacing the whois software with the OSI Directory Services was expressed.
The roles of the ripe and ripe-org distribution lists were clarified. The ripelist was a public list while ripe-org would be restricted to active participants.All messages sent to the ripe list would be forwarded to ripe-org. Thus there would be no need for anyone to subscribe to both lists.
The discussion included a progress report on each task, a presentation on the mapping of IP across PDN's, some comments on running name servers and some remarks on a European strategy for name servers were given.
3-1: Namespace administration database. Francis Dupont proposed a format for the Namespace Administration Database. It was suggested that this should include the duties of the administrators. Some problem areas were identified, such as EARN not using domains in the expected way.
3-2: Coordination of secondary name servers. Two needs were identi- fied: - contacts on national, European and US level; - a strategy... Mats Brunell opinioned that name servers were a national concern and thus nameservers should be on the national level. Contracts for secondaries should be negotiated. Daniel Karrenberg gave some comments based on his own experience. Firstly, having too many name servers produced very real management problems. He also described the name server set up at CWI which used two physically independent workstations with stablised software and no users. One of the best solutions would be to have a primary at the national level, and secondaries at the European and US levels.
3-3: Framework for secondary nameservers. INRIA proposed using RIPE to maintain contact with name servers in the United States. Discussions on placing a legal root server in Europe were slowly progressing.
3-4: Recommended Domain Name Server software. There was a problem with some bind software that resulted in a name server appearing to die. This was due to queries being disregarded for periods of some minutes during the transfer of zone information to other name servers. Bind 4.8.2 fixes this bug and also supports negative caching so that frequently asked but wrong questions would not require a return flight across the atlantic. Domain Name Server implementations are still evolving and still leave a lot to be desired.
The software will be available as part of the RIPE toolbox and be available by ftp. The dumb Yellow Pages software uses far too much Intercontinental bandwidth.
It was suggested that RIPE should talk to Phil Gross, from the US Internet, about a formal proposal and paper stating exactly what is required. This work need not be done until after March 1990.3-5: Mailrouting and MX record Study. Daniel Karrenberg expressed the view that because of the current political sensitivity of this issue, the meeting should avoid discussing it until a more suitable time. Rudiger Volk volunteered to delegate someone to updating the database, although the opinion was expressed that this should be done by or in conjunction with local people.
Daniel Karrenberg thought it better to restrict the number of name servers as experience showed that manageability decreases as their number increases.A first draft of the strategy for Nameservers in RIPE was required in a short period of time. With similar priority it was thought that a discussion ought to be started within a short period of time on producing a proposal for an official European Root Server.
Progress of the tasks were reported with particular emphasis on the draft agreement.
4-1: Template for IP coordination agreements A draft agreement on line sharing was distributed, based on an existing EUnet/NORDUnet document. CERN would be asked to have their lawyers comment and perhaps improve on it. Everyone present was asked to read the text and comment on it and the text was examined in some detail at the meeting itself.
Mats Brunell expressed the opinion that it should be more strongly worded, and include the statement \free traffic flow allowed". In addition, he asked that members should not charge each other for use of internal networks. He also informed the gathering that RFC 1009 (Requirements for internet gateways) was being updated. It was thought that there may have been a necessity to define exactly what would allowed over a line. Special services, such as EUnet, would not be restricted but it is not clear whether DECnet, for example, would be allowed.
Additional agreements would be needed for backup lines too. Some discussion took place over the problem of what to do when bandwidth was actually used up, this continued into the area of clauses for the cancellation of agreements.
4-2: Formal contact with CCIRN and FRICC. If RIPE was to be a peer of the US Internet with full mutual interconnectivity, some mechanism would have to be developed for connection agreements. There was a feeling that bilateral agreements would be inappropriate for this. The priority of this task was debated, some thought that creation of a unified European network was far more urgent although it was pointed out that the US were already making plans for Europe and so RIPE ought to get involved right away. Problems were envisaged with the US regionals.
4-3: Formal contact with other organisations. Barry Leiner, of the Internet Activities Board (IAB), had proposed an ISO-IP project which NORDUnet would participate in. Mats Brunell proposed that other RIPE parties also takes part.
4-4: Legal status of RIPE. A strict definition of RIPE's legal status was to be deferred until after the RARE COA meeting the following week.
The next RIPE Meeting was planned for the 22-23 March 1990, at theNIKHEF in Amsterdam. The meeting would begin at 14:00 on the 22nd.
The actions most urgent requiring work were identified. A list of them follows: 1-1; 1-3; 1-4; 2-2; 2-3; 3-1; 3-2; 3-3; 4-1; 4-2/3.