Second European Operators Forum 15 July 1994 Chaired by Peter LOTHBERG Minute took by Marc PICHON Hosted by TRANSPAC FRANCE Montparnasse Tower Paris Dear EOFers, This is the minutes for the Paris eof meeting, 15 July 1994. Marc ============================================================================= Marc PICHON pichon@rain.fr Phone # 33 1 40 75 07 11 TRANSPAC Fax # 33 1 40 74 02 87 206, rue du faubourg Saint Honore 75008 Paris, France ============================================================================== PARTICIPANTS Name Organization Email Peter LOTHBERG PL EBONE roll@stupi.se Bernard TUY BT RENATER tuy@renater.fr Stephan BIESBROEK SB BELNET stephan@belnet.be Christophe CHAILLOT CC France Telecom chaillot@cedre.france-telecom.fr Marc PICHON MP TRANSPAC/RAIN pichon@rain.fr Van Phuong NGUYEN VN TRANSPAC/RAIN nguyen@laba.rain.fr Nigel TITLEY NT British Telecom NIGEL.TITLEY@AXION.BT.CO.UK Jean-Michel PLANCHE JP OLEANE jmp@oleane.net Keith MITCHELL KM PIPEX keith@pipex.net Frode GREISEN FG EBONE frode.greisen@uni-c.dk Alain PEGORARI AP France Telecom pegorari@cedre.france-telecom.fr Gilles ANTOINE GA TRANSPAC antoine@rain.fr 1 ADMIN 1.1 EXCUSES Excuses were received from Daniel KARRENBERG and Per Gregers Bilse. 1.2 MINUTEMAN Marc PICHON volunteered !!! 1.3 APPROVAL OF THE PREVIOUS MINUTE The Lars-Johan LIMAN's minute is approved. However, the word "NATIONAL" (chapter 1.6) should be removed. 2. TRAFFIC AND ROUTING EXCHANGE MODELS BETWEEN PROVIDERS After KM gave his talk(Emailed by KM the 12th of July 1994 on the eof-list) here are the comments made by the participants: FG: In EBONE we have such agreements but they could be a problem because of: - non English speaking people - the juridic aspects, however, they are still useful. PL: The scalling process is very important, we need guide, samples. PL: What does mean "Non-Transit" ? PL: How do we troubleshoot the network ? PL: Email is not an Internet Service. PL: The two parts should optimize the backbone usage. PL: CIX is not relevant but as an example. PL: No need to differentiate NATIONAL and INTERNATIONAL if we define what transit means. PL: How do we proceed to fix problems between operators? FG: Could we have example of real agreements that have been writing down, Are they public ? KM: It is difficult to describe every problems but we could describe a few as transit, third party, back up, .... So, it will be possible to solve local problems using standard. SB: Neutral Interconnection could solve most of the problems. KM: We need a document defining relations in neutral interconnect. How do we define a route server ? KM: Neutral Interconnect are not described in this document. Who pays for the facility management ? PL: By management we mean the guy who resets the router or take care of the modem. In Europe, the Phone Companies take care of the modem. KM: We expect a new version of this document before 6 weeks from now. Conclusion: KM and PL will work on this document. 3. Connectivity from Europe to other continents via the US. Canceled due to the absence of Bernhard Stockman. 4. WHAT ROLE WILL THE US NAPs PLAY IN THE GLOBAL INTERNET The NAPs (Network Access Point) will be operational by November 1994. The NAPs are sponsored by the US government but the customer will have to pay to be connected. Three NAPs are forecast: a) in California using SMDS technology b) in Chicago using ATM technology c) in New york using FDDI, Bridge and Ethernet technologies. At this time we do not know if the networks DOE and NSI are going to get a connection to a NAP and to which NAP. As the NSF Backbone will disappear soon the problem is now: in order to get full connectivity to the rest of the world to which NAP should we connect? A rule of thumb will be to follow the same politic as the big networks. 5. ROUTING REGISTRY ARCHITECTURES Canceled due to the absence of Daniel KARRENBERG. 6. D-GIX ISSUES, A COMMON INTERFACE BETWEEN NSPs. At this time the only routing technology that runs fine is based on routers. But now, there is a need for a more efficient and flexible technology, in particular the D-GIX. If you have a global exchange routing point you can change from one service provider to the other without spending too much money. Also, the route service has to act on the level two to allow source routing . But it is possible to overcome this drawback by putting a router on an ethernet with the ability to talk to other routers on the same ethernet. That way is used in Stockholm with five differents services providers and it is planned to be the same in Paris. The drawback of this configuration is that many people have to be contacted for any change in the configuration. 7. CIDR AGGREGATION OF LARGER BLOCKS The CIDR aggregation had permit to reduce the size of the routing tables. These aggregations were usually made by the regional service provider but now the internet is facing a new increase in the size of the routing table. To be more efficient this aggregation should be done at a higher level. For example, we know that the networks numbers from193.x.x.x to 194.x.x.x are dedicated to europe and on the other end the networks numbers from 198.x.x.x to 199.x.x.x are dedicated to the US. So, it seems to be possible to announce these routes with only four routes, but how ? The problem is under discussion but we need good suggestions. The route server cannot do any aggregation. We could renumber and by this way have as much routes as AS number (around 600). 8. EOF CHARTER SB: The EOF report should be sent also to the ripe-list to overcome the limitation of the eof-list. PL: The Internet is not only a US infracstructure so the EOF should be the European equivalent of the IEPG. FG: The EOF should act as a European part of the IEPG. FG: The words "Internet-related networking services" in the chapter "Open Membership" should be replace by "ip-service". FG: Also the sentence with "... touristic or educational interests ..." should be rephrased. FG: It will be better to separate the historical notes. NEXT AGENDA (needs suggestions) Election of a new Chairman. PL will stay chairman until the next meeting, meanwhile every proposition for a new chairman are welcome. TROUBLE TICKET HANDLING Touble management Procedure, use AS number, use prt.traceroute, ... INTERNET MAP How to draw the world starting from a router AS table. Next meeting 12 September 1994, from 9 to 12,30 in Lisbon just before the next ripe meeting.
I note in the EOF minutes the following phrases:
For example, we know that the networks numbers from 193.x.x.x to 194.x.x.x are dedicated to europe and on the other end the networks numbers from 198.x.x.x to 199.x.x.x are dedicated to the US. So, it seems to be possible to announce these routes with only four routes, but how ?
I don't want to reduce your enthusiasm, but I would like to observe that we currently have over 25000 routes. Reducing 25000 to 4 is an interesting objective, but reducing it to e.g. 250 would already be a very good result. Or even 2500, for that matter. One of the reasons why I do not believe in the "reduction down to 4" is that it smacks of monopolies, implies a very unusual amount of coordination. It is perfectly sensible for an European network provider to manage its own international connectivity, without going through a particular "common fat pipe". This is not only natural, but a very sound competition strategy. And it does imply that this provider will announce its own reachability lists, without aggregating them under the continent's prefix. Christian Huitema
Christian Huitema <Christian.Huitema@sophia.inria.fr> writes: * I note in the EOF minutes the following phrases: * * >For example, we know that the networks numbers from 193.x.x.x to 194.x.x.x * * >are dedicated to europe and on the other end the networks numbers from * >198.x.x.x to 199.x.x.x are dedicated to the US. So, it seems to be possibl * e * >to announce these routes with only four routes, but how ? * * I don't want to reduce your enthusiasm, but I would like to observe that we * currently have over 25000 routes. Reducing 25000 to 4 is an interesting * objective, but reducing it to e.g. 250 would already be a very good result. * Or * even 2500, for that matter. * Christian, I would like to point out that the 25,000 figure is incorrect. As of last night the number of routes seen was just under 20,000. Whilst I admit in the context of this dicussion the order of magnitude is the same it is very important NOT to mis-quote figures in the current climate of discussions. If you want to check into the figure at any time please look at ftp://ftp.ripe.net/cidr/stats/Table-History * One of the reasons why I do not believe in the "reduction down to 4" is tha * t * it smacks of monopolies, implies a very unusual amount of coordination. It * is * perfectly sensible for an European network provider to manage its own * international connectivity, without going through a particular "common fat * pipe". This is not only natural, but a very sound competition strategy. And * it * does imply that this provider will announce its own reachability lists, * without aggregating them under the continent's prefix. * With this I totally agree. Not in the sense of monopoly building but purely from a practical engineering perspective. The idea of super aggregates have been talked about many times within the CIDRD group at the IETF. My personal view is the topology is so diverse and complex that whilst you can perhaps announce these aggregates towards a few places the effort of finding these places and keeping track far outweighs the gain which will only be seen at certain points and not widely. If my memory serves me right the CIDRD group had decided to in the "near term" at least, concentrate on acheiving more aggregation at the provider level where there are still significant gains to be made and I would urge the EOF group to follow in this effort for now at least. Estimations for aggregation at the AS level are also done every night by the way ftp://ftp.ripe.net/cidr/stats/AS-agg.* and there appears (remember this is just an estimation) shows even at this level we could back another 5,000 routes tomorrow. --Tony.
participants (3)
-
Christian Huitema -
pichon＠rain.fr -
Tony Bates