to BGP - and BGP already is quite an expensive and high-overhead protocol in terms of communication (compared to link-state anyway, which is O(1) messages per link for any link-event; path-vector is something like O(# simple-paths between link-event and receiver), perhaps more, for worst case).
Oops, path-vector is worse, forgot about all the destination, so at least O(# simple paths between link-event and receiver * the number of best-paths to destinations beyond the link event). Simple-paths is a function of the size and complexity of the network.
Tx-ing every neighbour AS best-path will add a further multiplicative, variable factor to that overhead, that also scales up with the network complexity).
Is MED really _that_ useful a metric that we should incur significant costs on pretty much _all_ BGP processing? There is surely some cost-point beyond which it would be better to just give up on MED as currently specified?
If instead we just constrain MED (and BGP metrics generally) just a little bit, none of these expenses are required.
regards, -- Paul Jakma [email protected] @pjakma Key ID: 64A2FF6A Fortune: Humor in the Court: Q: What is your relationship with the plaintiff? A: She is my daughter. Q: Was she your daughter on February 13, 1979? _______________________________________________ Quagga-dev mailing list [email protected] https://lists.quagga.net/mailman/listinfo/quagga-dev
