Daniel Karrenberg of RIPE NCC says that the IP number registries are seriously considering issuing Class A space in a classless manner. AGIS is willing to consider this.
The AGIS routing policy filters /24 for 192 - 205
The AGIS Peering policy would keep another AGIS from coming along. How do you justify that? AGIS is doing this for router survivability. The availbility of colocation space also affected this.
The policy does define what AGIS sends and what AGIS accepts.
What are people doing about IP Multicast?
AGIS is not supporting IP Multicast, but they are willing to consider it. IBM is just now begining to plan IP Multicast for deployment in the first quarter next year.
ANS is not doing it and Sprint will consider implementation towards the end of 1997.
Kent England asks how bad the default routing is at the NAPs. Sprint advises that they have had a major problem with this with one peer. They were able to make adjustments to compensate for this. ANS had a similar problem and just turned off the peering session. IBM has not had this problem, but detecting it is not easily detemined by IBM's tools. AGIS depends on reports of asymmetric routing. AGIS has had problems at the MAEs with folks that are not peers pointing routes at them.
Bill Simpson asks what is being done to work together towards getting bigger and better pipes (like across the Atlantic). Curtis says that cooperation is not realistic to do this.
Enke Chen wants to talk more about defaulting at the Interconnects. He says that this can be solved by Mac-Layer filtering at the router or port filtering on the gigaswitch. He suggests that loose-source routing be enabled for all peers. Curtis says that ANS will inform the offending provider that they will shut down peering and then they will do it.
Alot of private interconnects are coming up. Will the MAEs still be viable? IBM favors the public exchange points. Curtis says that the purpose for creating the private interconnects was to remove MAEs traffic, not eliminate the MAEs. Sprint agrees with ANS on this.
Alan Hannan asks What is a "broke" Internet? Is the Internet broke? Will it ever be broke? AGIS says that it will never be totally broken. IBM says that for the Internet to collapse, it would have to be a chain reaction. He considers this unlikely. Curtis says that there have always be claims that the Internet will die, but it hasn't. However, there is a possibility since there only a few sources of router software. Even if a router bug takes the Internet down. Recovery should be pretty rapid. Sprint sees that the effort to continue to deliever services will continue stress the environment, but with appropriate responses from the providers, this stress will only cause the Internet to grow and get more tolerant of broken-ness.
The Internet is pretty much a single source of product. There is definately a need to make the network more multi-vendor-based. ANS agrees. Sprint is constantly looking for new sources.
What about dealing with hostile attacks? IBM has been doing source-address filtering for 1 1/2 years. ANS is adding source-address filtering at this time. Sprint is not doing any IP-address filtering. They are investigating this at this time. AGIS is considering how to add IP-address filtering.
Sprint is not accepting new peers at this time, but will be at sometime in the near term. They are re-evaluating their policy and once that is complete they will be able to consider more connections.
AGIS does require prospective peers to commit to a certain level of infrastructure before they will peer. As for cross connects that the NAPs themselves, that is really a NAP operator issue.
Why the DS-3 rate? Why not a 10Mb/sec? Sprint says that peering should be something that is roughly equal between the peers. IBM does peering on a case-by-case basis.
Bill Simpson says small ISPs should seek peering with those that have "good" policies. He also asks what the providers are doing to encourage local exchange creation? AGIS says that they are in favor of their development, but will only participate when it has a specific benefit to them. IBM will not go to a small exchange point in the US unless they can sell transit there. Outside the US, different rules apply. Curtis agrees with IBM and will also participate at exchanges with the other major providers. Sprint agrees with ANS.