> Most likely, this would imply routing towards the specific > exit based on the packet's source locator in addition > to the destination locator. This is not part of the base > ILNP spec as it exists today.
> well, does it need to be? Good question. I'm certainly happy to add some MAY language to the ILNP base spec about an originating node having multiple known Locator values indicating the node's preferred egress by thoughtful selection of the Source Locator value. I am very reluctant to require routers to use Source Locators as part of their forwarding/routing procedures, as it adds complexity, is not fully backwards compatible with installed IPv6 routers, and because this kind of thing seems a lot like a differentiating feature rather than a required basic capability. > If the host puts a src-locator on the packet it creates > [then] it's up to the network/host folks to agree to > src-route or agree to get overwritten. In the end, routers along the forwarding path for any IP (or ILNP) packet make a myriad of decisions based on both global policy, originator decisions (e.g. which destination to put into the packet if there are several), and local policy (e.g. hot potato routing). While an originating node gets some input into that distributed computation/procedure, the node with the packet at a moment in time inevitably has the ability to make its own decision (some of which might result in a packet taking a sub-optimal path or being lost). > (Is the 'TE' at the whim of the host or the > network? This sounds like shim6 discussions, to me) Originator can indicate a preference, but originator preference (in practice, no matter what any RFC says) will not over-ride contrary local policy in some router along-the-path as it forwards that packet. Yours, Ran _______________________________________________ rrg mailing list rrg@irtf.org http://www.irtf.org/mailman/listinfo/rrg
