Implementation of RIPE-181++
Tony Bates
Thu Sep 22 18:01:03 CEST 1994
Marten Terpstra <Marten.Terpstra at ripe.net> writes: * * Laurent Joncheray <lpj at merit.edu> writes * * Hey dudes, any winner so far? I reaaly like to see the algorithm * * in order to be able to implement something. So far i got half of the ans * wer * * from Marten and a complete (but wrong) answer from Cengiz. * * Enjoy... * * Laurent * Okay, I am jumping in a little. However, I will try to explain it as I tried to write it in the document. Basically, the as-in line is always the superset of the interas-in policies. Lets take each one in turn of possible problems. I have written in sort of simple terms as well. 1) You have policies in interas-in not described in as-in. They are ignored. To work it out....it is the difference of set of interas-ins minus the as-in and what is left is ignored. 2) You have more policy in as-in than described in interas-in. The polices not specifically defined in interas-in but mentioned in as-in are deemed equal across all interas-in peers/links. To work it is the difference of the as-in minus the set of interas-ins and what is left is shared across the links/peers. This implies adding to basic list generated for interas-in lines. 3) You have something like AS granularity in as-in and some network granularity in interas-in. i.e. as-in: ASX 100 AS2 interas-in: ASX a b blah {15.0.0.0/8} interas-in: ASX c d blah {10.0.0.0/8} You must always apply rule 1 above. First make sure 10.0.0.0/8 and 15.0.0.0/8 are in AS2, otherwise they are not accepted. If they are then make list such that 15 is best a-b anf 10 best c-d and the rest of AS2 is equal through both links. --Tony. -------- Logged at Thu Sep 22 19:04:12 MET DST 1994 ---------
[ rr-impl Archive ]