I am working on Quagga, on behalf of Euro-IX, to allow for future growth and other developments specific to the needs of a Route Server.
The discussion in 7.4.2 is incomplete. It may be true that traffic is 80:20 in favour of direct bilateral peering (Method A). However, that does not mean that 80% of the peering connections are bilateral peering -- the distribution of traffic amongst peering connections has the usual long tail. Numbers vary, but a large exchange may have 50% of its clients connected to the route server for most (if not all) of their peering connections. Section 6.6 discusses Proxy Signing. Where a route server is not inserting its own AS in the path (Method B), it is acting as a proxy for each route server client. Given a certificate and private key from each client, the route server can rewrite the most recent signature. So, and please correct me if I have this wrong, the case is covered ? Incidentally, the discussion notes that router servers that insert their AS in the path (Method C) are rare. I'd put it more strongly than that, and relegate the case to "footnote" status. With such a route server the IXP has all the appearance of a transit provider (at least at the route level), so would be trivially supported -- at least until some mechanism is devised to validate the AS Path and the path taken by packets against each other. Chris _______________________________________________ sidr mailing list [email protected] https://www.ietf.org/mailman/listinfo/sidr
