First, while it is interesting to look at YANG models, what we need to be thinking about in i2rs is information models - and how well different data modeling languages meet those needs.
More comments in-line (ruthlessly cut) On Mar 20, 2013 2:02 PM, "Russ White" <[email protected]> wrote: > Rather than asking how to model a RIB, what if we tried to model the > information needed to actually forward a packet at layer 3? In other words: > > - These are the tuples needed to build a solid routing table entry from > which the device can build forwarding table entries > - Add in the pieces we need for "policy," particularly the idea of > preference (called admin distance in cisco, for instance) [Alia] I don't think that the forwarding info is quite the set we need, but I would really like to see even a minimal information model for what the RIB service info model should look like. For instance, in addition to admin-preference, we need the multiple route-tables, ways of putting traffic in other than by just dest IP address, what events can be registered for, etc. > Is there anything else we would need? > > The other option is to simply realize there's no "right" model here. > Just choose one that works, and move on with life. [Alia] yes, get a small set of sufficient info to meet a basic set of use-cases and define & use it. It can always be extended later if and as necessary. Alia
_______________________________________________ i2rs mailing list [email protected] https://www.ietf.org/mailman/listinfo/i2rs
