Rob Blokzijl welcomed the participants.
Apologies from:
Bernhard Stockman found himself taking the minutes.
Papers to be discussed during the meeting was distributed:
The arrangements for the Tuesday evening "rijstafel" were described. It was pointed out that "Udang Besar" was missing.
The agenda was approved. Some changes maybe necessary due to working group sessions to be decided on later.
The minutes from the last RIPE Meeting were approved.
Rob Blokzijl
Tony Bates and Marten Terpstra
Mats Brunell
Daniel Karrenberg.
Thomas Lenggenhager and Bernhard Stockman
Willi Porten
Matti Rendahl.
John Seymour
Bernhard Stockman
Approval of NCC activity plan. The RIPE NCC activity plan had been distributed and discussed at the last RIPE meeting and was now formally approved. The activity plan was presented to RARE which agreed on the content.
It was noted and discussed that the selection committee for the RIPE NCC manager, staff and location will be made up of RIPE and RARE individuals and that no funding body would be allowed to have a say in these matters.
Approval of NCC organizational framework. RARE will provide its legal framework but will not fund the NCC from its own budget. Money could however pass through RARE dedicated for the NCC. RARE will provide the fiscal framework for the RIPE NCC.
RARE decided at the last CoA meeting to install a task force to study an organizational unit for network management. The task force will report at the next CoA meeting in October.
Rob Blokzijl is to determine the best way to create a chain of command between the NCC Manager and RARE. One recommendation was that the NCC manager report directly to one CoA member.
Bernhard Stockman asked if there could be any problems with including commercial IP providers in the NCC coordinating activities. Rob Blokzijl explained that known commercial providers will be invited. RIPE's ToR clearly include non-academic networks. Rob Blokzijl will stress this in his upcoming presentation at the INET '91 conference which many such providers are expected to attend.
The paper on the set up of the RIPE NCC was shortly discussed and approved in principle. Rob Blokzijl will distribute the document on the ripe mailing list for further commenting.
Some parties interested in funding were mentioned. Some of those could be hallmarked as "external parties" (outside the usual RIPE/RARE scope). Don Stoikvoort put forward that this funding by "external parties" is very enjoyable, but that it would be a very good thing if the parties/networks directly interested in the NCC's activities - and benefiting from those activities i.e. the (inter)national researchnetworks and networking organisations, like NORDUnet, EUnet, JANET, DFN, SURFnet, HEPnet, CERN, etc. - would contribute financially to the NCC right from the start. This could preferably be done using the "funding via RARE" mechanism. Don Stikvoort made an appeal to those present to organise this kind of funding. e.g. via their RARE representative (if available).
Ruediger Volk and Daniel Karrenberg agreed it would be a good thing to get financial involvement from the participating networks right from the start. Rob Blokzijl expressed as his opinion that at this moment the most important thing is to obtain the funding for the NCC.
The solicitation for the RIPE NCC will be announced on the RIPE and RARE CoA mailing lists. A small group consisting of the RARE EXEC and RIPE people will review the answers.
The solicitation will be sent out before end of June 1991. The deadline for answers will be around mid-August. The selection should be ready a few weeks after this.
Western Europe The IP over IXI connectivity will be treated in the session for this topic and the results reported the Thursday plenary session.
Central and Eastern Europe The present central European representatives reported from their respective countries.
Poland Not much international connectivity, only two lines. Some national lines exist but due to the infrastructure only 9.6 Kbit/s available.
9.6 Kbit/s EARN Univ. of Warzaw
Copenhagen o----------------------------o Info.Centre.
/ \
/ EARN
Physics / \
Dept. o o Dept. for
Warzaw \ Cybernetics
DECNET Cracow
\
CERN o---------------------------------o
9.6 Kbit/s DECNET Jagelonian
Univ. at Cracow
The line from Cracow to CERN will run IP very soon. Within Warzaw SLIP is used between PCs. There is hope to use SLIP to between Warzaw and Copenhagen. There is also a 64 Kbit/s planned satellite link between Warzaw and Stockholm. CISCO routers have been ordered but are delayed due to CoCom restrictions.
Hungary A X.25 network is installed and run by the Hungarian PTT. Most universities are connected, mainly using 9.6 Kbit/s lines. There are a few leased lines and for private use only. On the international level the X.25 network is connected to corresponding West European X.25 networks. A 9.6 Kbit/s leased line is installed between Linz and Budapest. An other 9.6 Kbit/s leased line is ordered between Budapest and CERN which will run IP.
IP connectivity exists within isolated areas within Hungary. The Hungarian PTT can offer 64 Kbit/s lines within Budapest only. University sites in Budapest plans to interconnect using FDDI. There is a need for one or two 64 Kbit/s multiprotocol leased international lines connecting to West Europe EARN and IP networks. There have been discussed connections to Trieste and Wien. CISCOs impossible to get due to the CoCom restrictions which really is a problem.
Czechoslovakia Dial up connectivity between Bratislava and Amsterdam using trailblazer modems. A dedicated 9.6 leased line between Bratislava and Wien is ordered. The line will be multiplexed with 50% for IP and 50% for X.25. There is no X.25 network within Czechoslovakia. Discussions have started between two US companies and the Czech PTT to install a national X.25 network. There are however problems in this due to CoCom restrictions.
A 9.6 leased line between Prague and Linz running BSC for EARN. The line is paid by the Czech government for EARN services. IBM EEAI includes a planned line from IBM office Prague to IBM office Vienna. There are seven EARN nodes within the Prague area. Two 9.6 leased lines, from Prague to Bruno and from Prague to Bratislava, are planned. A project with a the French government aims at installing a 64 Kbit/s backbone interconnecting Prague, Bruno and Bratislava. This backbone will however not be in place before around one year. Another joint project between Germany, Poland and Czechoslovakia is started to get connectivity to libraries.
Jugoslavia A 64 Kbit/s leased line connects Ljublajana to Wien. A 9.6 leased line for EARN is installed between Belgrade and Linz. Around 300 sites are interconnected by the national public X.25 network. There are two systems running X.400 software (the slovenian PTT and the UNACnet). There are some isolated LANs. Two sets of FDDI networks are expected in Slovenia. The EUnet backbone site in Ljublajana is also connected to the UNACnet. There are plans to use IXI for IP connectivity. The numbers of SUN systems are increasing fast.
Daniel Karrenberg expressed the concern that we should try to help these countries by providing PCs and public domain routing software while the CoCom restrictions prevent dedicated routing equipment to be delivered. This could be done as a project.
Tunisia Tunisia have recently been connected to INRIA in Sophia/Antiopolis via X.25. They have a fully installed IP connectivity including a name server and full connectivity to NSFnet. Currently the useful bandwidth is 2.4 Kbit/s. On the INRIA site the line is terminated in a CISCO and on the Tunisian side at the IRCIT institute in Tunis to a VAX/VMS system using the Wollongong TCPIP software.
At the technical session on Wednesday, convened by Milan Sterba, the subject of North African connectivity will be treated and a paper produced clarifying the subject.
ESAnet ESA-internal IP connectivity has been set up between the major establishments and to NASA-IP networks.
ESA - as an international organisation - is currently studying concepts to interconnect this internal network to the worldwide open Internet, both within Europe and in the US. This includes issues such as access points to national parts of the Internet in one or multiple countries, routing between those national networks and the ESA-IP network, security, etc, as well as the setup of domain name services and having ESA registered under (a) top-level domain(s). Advise was sought by Walter Dillen to which top level name server domain ESA should belong.
So-far, one 64 kbps link has been configured from the ESA-ESTEC establishment (NL) to NLNET via CWI-Amsterdam, which provides limited connectivity from within ESA into the Internet.ESA is participating in IXI and has an IXI access point at ESTEC. ESOC is connected to DFN/WIN but not yet using IP on this. ESRIN in Frascati, outside Rome, is just a few hundred meter from a GARR node.
Presentation Milan Sterba and Marc Sheldon gave an introduction on the use of IP over ISDN. The overheads were copied and distributed to the participants.
Follow up There is no real needs to have any actions in this area for the time being as there is not enough connectivity in Europe yet. However there is a need to be informed on what is going on. Marc Sheldon will keep track of what is happening.
Following groups will meet at this RIPE Meeting.
Will be treated in a Wednesday session and reported at the Thursday plenary. A paper on the latest DNS host count was distributed by Arnold Nipper.
Will be treated in a Wednesday session and reported at the Thursday plenary.
Will be treated in a Wednesday session and reported at the Thursday plenary.
Report Daniel Karrenberg circulated a written report including database statistics which was discussed. The report is available from the RIPE Document store.
Due to the large amount of updates and additions to the database there is no time to implement the code necessary for including the new objects. This will be delayed until the RIPE NCC is in place and can undertake this task. Yves Devillers proposed that new objects should be announced nonetheless for information purposes and to start gathering information. Daniel Karrenberg will make templates available time permitting. [Ruediger Volk has voluntered to help with this after the meeting]
The standard database software distribution is still available. This does not include new consistency check tools and database code to handle domain objects. The new version of the database software can according to Daniel Karrenberg only be delivered on a beta test basis. There exist currently no complete software distribution with the necessary makefiles etc. Until this is implemented the software will not be made public available.
Sylvain Langlois, EdF, have earlier expressed interest in undertaking the implementation of an X.500 interface to the RIPE Database. No information on the progress was available at this meeting.
Two papers have been produced by the management task force:
Both papers were discussed at the previous RIPE meeting and have now been accepted as formal RIPE recommendations.
Security issues should be addressed. CERT should be taken into Europe. RARE WG8 have been started for this items. RIPE related actions should be coordinated with RARE WG8.
END OF DAY 1
DAY TWO FULLY DEDICATED TO WORKING GROUP SESSIONS
DAY 3:
To make sure that most participants should be present during this point, it was moved to the beginning of this day. The 10th RIPE Meeting will as decided earlier be at CERN, September 25 to 27, 1991. The 11:th RIPE meeting will be in Amsterdam, January 20 to 22, 1992.
It was suggested that additions and changes to the DNS should be done in reasonable time. One week in advance could be regarded as reasonable.
Piet Beertemas paper on problems and recommendations on installing and running a name server was studied. The paper is available with anonymous ftp from nic.eu.net. This paper should not be regarded as the earlier proposed DNS starter kit but as a tutorial "do and don't".
There is a recognized need for a starter kit on DNS. There is also a need to document other similar paper. A document is hence needed which contains pointers to other relevant starter kits.
There were a discussion on the location of top level name servers. Organizations should be found willing to host top level servers. There is also a problem with domains not having a secondary server in the US.
Yves Devileers raised the issue that those managing a top level domains should process requests with minimal delays. As a rule Yves proposed that one week would be acceptable while longer periods, like two months experienced in some cases, is unacceptable.
The session was based on a draft document describing which parts of a network infrastructure of an organisation participating in RIPE should be documented in which way. I.e. which objects belong to a local network map and which ones belong to a RIPE map.
There was soon a conclusion, that there are different needs for maps which will result in different types of maps needed:
An example for the Topology Map is the RIPE map, for the Policy Map e.g. a map with the virtual circuits for IP over IXI.
The draft document will be progressed to define/recommend the contents of Topology maps, for both RIPE and local levels. For maps of other types we most probably need more experience, i.e. we need first examples to review and discuss.
In addition to a definition of the contents of maps, there is the need to have a common set of symbols used in all the maps. Bernhard Stockman pointed out, that within the IEPG this need was expressed too. Coordination with the IETF has to be sought.
As a first step, Hank Nussbacher will put together a list of locations where everyone can get the newest versions of network maps, especially the ones of the regional networks in the US. Based on that a review of the currently used symbols should take place, to come to a 'standard'.
A longer term subject is the need of a database format usable for automatic/manual map generation. Some progress can perhaps be made together with the network management systems requiring similar data.
James Barr volunteered for the job as the convener of the mapping group. Bernhard Stockman will send the necessary information.
Central and Eastern Europe connectivity session report
Overall networking situation As for the Central and Eastern European countries representatives from Czechoslovakia, Hungary, Poland and Yugoslavia took place in the session.
In the first part of the session a brief summary of existing and planned international links to the Eastern and Central European countries has been done as well as an overview of existing network cooperation projects.
It has been decided that a RIPE working document will be produced summarizing the actual status of networking in these countries. Borka Jerman-Blazic, Jan Gruntorad, Nandor Horvath and Rafal Pietrak will furnish a national report on situation in their respective countries to Milan Sterba who will summarize and add the international connectivity and projects staff. Milan Sterba will also set up, maintain and publish a list of persons responsible for academic networking in these countries with their postal and e-mail addresses and other pertinent information.
All the countries involved are facing a similar problem of limited networking resources and an increasing connectivity need inside as well as outside the country. As different networking activities are emerging in parallel in these countries (EARN, EUnet, Internet, HEPnet, DECnet, COSINE Project ...) there is a pressing need of efficient sharing of both national and international infrastructure. Many of these countries are also limited in technical connection possibilities to low speed links and do not have easy access to high-tech multiprotocol routing solutions. It is often easier for these countries to get some suitable international connectivity than to distribute the access to it nation wide.
The EARN - TCP/IP line sharing problem The problem has to be tackled in both political and technical ways. We suggest RIPE officially approaches European networking organizations (especially EARN) to invite them to cooperate on improving Central and Eastern European countries connectivity with special attention to line sharing in order to obtain from these organizations an official statement on line sharing.
From the technical point of view almost all countries face the same problem of using existing EARN lines, financed by their governments, for NJE as well as for IP and maybe other types of traffic (DECNET). The problem is the same for international as for national leased lines. A short term technical solution is for this problem is needed.
The austrian EARN organization (AEARN at Johannes Kepler University in Linz) is very cooperative in solving the EARN-TCP/IP line sharing problem.
A small technical group has been formed and has evaluated all the possible technical solutions of the EARN - TCP/IP traffic line sharing problem. A working document has been produced and will be presented at the EARN technical committee by Guenter Schmittner. Milan Sterba will continue to be the focusing person for gathering information about technical solutions of these specific line sharing problems.
In spite of these interim solutions the long term solution is NJE on TCP/IP on the VM/SP itself and an IP router usage. RIPE could help deblocking the situation with IP routers (cisco) licences.
Better representation of Central and Eastern European countries in the RIPE database All Central and Eastern European countries are strongly encouraged to register all pertinent information to the RIPE database (even about not connected IP networks). Updates from multiple national sources will be accepted, although there should be one monitoring person per country to check the national part of the database (maybe the top level domain technical contact).
These countries need a cookbook about the RIPE database and other network information and registration procedures (structure of the database, how to do updates, how to get IP numbers, how to register top level domains, ...). This information should be easily available even without IP connectivity.
Summary Problems - need for short term, cheap line sharing solutions (sharing NJE, IP and possibly DECNET and X25) in order to share international and also national leased lines. - problem with getting licences for IP routing equipment - lack of information about the network registration and information procedures
Though a European IP backbone with a common routing policy and common operational support would be very much prefered, the current situation does not yet allow for this. Despite of organisational and operational diversity of the current European Internet scenario there is a possibility to have a well coordinated routing core.
This RIPE coordinated routing core has to accomodate the differing policy requirements of the members contributing to the European level Internet; this currently precludes the use of an interior gateway protocol to create a common routing scheme. It is necessary to structure the routing core into several autonomous systems interworking by exterior gateway protocols and thus implementing policy based routing. This approach actually is already in use; but the current setup was derived by ad hoc decisions, has no overall design, and is not documented.
To improve on the current situation it was identified that the RIPE database needs to be extended to document routing policies; this asks for records to describe policies as implemented on the routers; it also will be necessary to define "privileges" and authorities that can grant them. Privileges will be represented by tags used in the "connectivety" field of the records.
RIPE is not encouraging restrictions, but known restrictions should be visible for those setting up routing rules. Each organization should give an explanation for their tag. Willi Porten will collect this information. It is expected that a collection of such data will show the need for some refinements of the database and improve understanding of policies and requirements, thus starting a few iterative steps to evolve documentation, implementation, and understanding of applicable routing policies.
For doing a proper design for the coordinated routing core the WG suggested to consider using BGP as the routing protocol. For this reason the WG proposes an BGP pilot to be undertaken and for this reason seeks volunteer sites. There is a need to investigate not just BGP but also to investigate the interaction between BGP and other routing protocols. Those interested in participating should contact Willi Porten. Willi will chose the BGP pilot sites.
There could be made a small budget available for this pilot from RARE. RARE could dedicate some funding for well defined projects. This should be investigated.
Jean-Michel Jouanigot will write a draft paper on policy based routing.
Regarding languages to describe routing policies the WG favours to look into applicable work of US parties over starting it's own effort to create such a language; this is not in conflict with the proposed approach to collect policy information in the RIPE data base (due to the envisioned rather informal ways to be used on the first iteration). Close cooperation with US colleagues on issues of policy based routing is considered very important; in particular it was suggested to organize a meeting at the upcoming IETF for a small group of people from the WG to draw on US expertise.
ESA have already tried BGP routing to NASA but with negative results. Hank Nussbacher was forced to use BGP to NYSERnet. This worked with CISCO OS version 8.1 but stopped work when 8.2 was installed. There is a need to document such experiences. Those having used BGP send your experiences. Willi Porten will put these together in a document.
Network monitoring and statistics gathering WG session
It was expressed the concern that restriction in SNMP access to RIPE central routers could be needed. There is otherwise the risk that these routers will be loaded with too much SNMP processing. Also there is the risk that lines will carry too much SNMP traffic. There is a need to have knowledge both on the SNMP load in the routers and on the lines. A possible way to reduce the line load is to have a local process performing the SNMP queries in a similar way to collect/statspy in the NNstat tool.
The WG discussed relevant polling periods. For long term traffic analysis giving the hour totals a polling period of 15 minutes was regarded as adequate. This will give the necessary resolution for producing diagrams on hourly usage. For peak behaviour there is a need for shorter polling periods. As CISCO currently have a 5 minutes traffic counter for each interface this would probably also be the polling period.
To facilitate for inter-NOC comparisons on network statistics there is a need to have a small set of commonly used tools. This could be undertaken as a project.
A draft document produced by Bernhard Stockman regarding the the definition of points of interest for statistical data gathering was discussed. The documents defines the lines and metrics to be gathered on RIPE connected networks. As the participants did not cover all the sites included, Bernhard Stockman will update the document with information from this session and send it out on the ripe-org mailing list for further comments.
There is a problem in gathering statistical data on IP traffic going over the IXI network. For some sites like CWI both IP and non-IP X.25 traffic is forwarded on the same interface which will give incorrect data if gathered. However at some points just the embedded IP traffic is being forwarded. For this reason the NIKHEF and ULCC IXI access points was regarded interesting. There is a need to have knowledge on the total IP traffic being transmitted over the IXI network and measuring that traffic on NIKHEF and ULCC will give a significant part of that traffic. There is currently no way to measure traffic on each virtual circuit on a CISCO interface connected to IXI. This could be done using the NNstat tool with a knowledge on which networks are sitting behind each virtual circuit.
James Barr gave a short presentation on the Monet tool. There have been a lot of bugs fixed. At the previous RIPE meeting it was decided that those interested could act as test site for the current beta-release. No reports from such tests were available at this meeting.
A public domain NMS named xnetdb was mentioned. Xnetdb is available via anonymous ftp from thor.oar.net (131.187.1.135) in the /pub directory.
Rob Blokzijl will hunt the WG conveners to have there sessions programs available one month in advance of the next RIPE meeting. The intention is to have the WG session programs included in the agenda.
Juha Heinanen gave a short presentation of the current status of the RARE WG4 CLNS pilot project.
The routing protocols involved are IS-ES which is to be used in NSFnet, IS-IS or dual IS-IS. The interdomain routing protocol is essentially BGP.
Applications that will be used in the pilot are X.400, X.500, VT and FTAM.
SunLink V 6.0 TP4 is 30 percent slower than TCP while the V 7.0 TP4 is 30 percent faster.
RARE RARE have been granted money from the CEC VALUE program. A major part will be used to enable the RARE WGs to have one meeting each year outside Brussels. Rob Blokzijl will apply money for one RIPE meeting as RIPE could be considered as a RARE WG. Some of the CEC money has also been used for RARE promotion material.
EEPG The EEPG group was dismantled at the last CoA meeting in Blois after having fulfilled its task to produce a report on the technical and organizational possibilities of creating an European Backbone.
As one result of this effort a database was installed giving information on international line. The maintenance of this database is for the time being performed by Bernhard Stockman but will be transferred to the RIPE NCC once installed.
IEPG IEPG had met in Paris on 21st and 22nd of May. One of the more concrete issues discussed at the IEPG/CCIRN meeting before last in Santa Fe was the stop of block allocation of IP addresses. This was questioned by Japan and RIPE because of the long turn around times and the need for local support. In Paris a paper was drafted on how to do block allocation in the future. A limit of 20 network numbers was suggested and that a check should be made that the allocated set was actually used before granting another small block of numbers. Any ideas about this should be fed back to Daniel Karrenberg.
CCIRN The CCIRN discussed at its last meeting in Paris the issue of Central European to USA connectivity. It was proposed that if an organization also cooperating with RIPE is installing connectivity that may cause problems, RIPE should be notified.
US networks have currently no permission to grant connectivity to Central and East European countries. Most agencies are supporting a relaxation of these restrictions but due to circumstances above there level of influence this is not possible for the time being.
The uttermost responsibility to block non permitted traffic to US networks are within the US networks but RIPE will cooperate to save meaningless traffic on the already loaded transatlantic lines. The problem will probably be resolved within the comming months.
CISCO. There is a need to establish formal contacts between RIPE and CISCO as RIPE connected networks mainly are using CISCOs. There could be the possibility of good business agreements and good support. Also to facilitate for the above proposed BGP pilot such contact would probably be beneficial. Rob Blokzijl will undertake this action.
Bernhard Stockman pointed out that after having scanned the previous RIPE minutes there are a lot of documents being proposed. Some of these documents have actually been produced and installed in the RIPE Document archive at nic.eu.net but a big part have either not been put online or have never been produced. There is a need to review the status of these documents. Bernhard Stockman will send a list of the relevant documents to Rob Blokzijl who will undertake the reviewing.
It was also pointed out that there could be a need to have the RIPE documents available via LISTSERV and BITFTP. This should be delegated to EARN which should come with a suggestion.
Rob Blokzijl thanked the participants and declared the 9th RIPE Meeting closed.
Rob Blokzijl
Jean-Michel Jouanigot
Daniel Karrenberg (?)
Thomas Lenggenhager and Bernhard Stockman
Hank Nussbacher
Willi Porten
Matti Rendahl.
Milan Sterba
Bernhard Stockman