Hi Brian,I think this topic is interesting indeed and i share Geoff feeling about complexity, but let me ask a bit more about this:
El 20/10/2005, a las 15:08, Brian E Carpenter escribió:
There's quite a lot in the draft about how preferences are communicatedand very little about where they come from. draft-ietf-shim6-arch-00.txtdoesn't help with this either. Somewhere, I think we need a listof possible sources of these preferences, and dynamic or static TE policyneeds to be in that list.
what sources do you have in mind here?- i can clearly see that local preferences like the local admin setting the local policy table to prefer some addresses over the others would make sense. This would basically affect the selection of source addresses and destiantions addresses used as ulids, as in the policy table of rfc3484 and hence the locator used until a failure occurs. In addition, a similar mechanism could be available for following locators (after an outage)
- however, there seems to be other players that may want to also influence in the preferences i.e. to be a source of preferences, in particular, the peer. The peer can convey preferences about which locators it want to use (this would only affect locators after a rehoming event, since these preferences are conveyed once the session is established so they are not good for selecting the initial ulids/locators) (i can see some value on this, depending on the scenario, as it has been discussed here)
Now, are you also considering here the isps as possible sources for preference settings on a shim6 host? i mean, do you think that transit provider should have the possibility to inform the shim hosts about what locators it prefers?
regards, marcelo
Some discussion of traffic engineering in the architecture document
would also seem necessary.
Brian