I assume the chairs will speak authoritatively on scope, but my current understanding is that this topic is still up for discussion.
On that topic, I'd like to point out that we have been vague about the definition of a "broker". I think we assume there is a centralized broker for each network element - because that lets us push the coordination problems (multi-personality arbitration etc) to somebody else. But we haven't explicitly talked about network-wide broker centralization, keeping the operational state in sync across multiple devices (to avoid loops and black holes), etc. Nor have we talked about whether the broker is really a single entity, or if it is distributed itself - and if that means we've really solved the problem of coordination (and sanity), or just avoided dealing with it here. I'd be interested in hearing thoughts, flames, etc about these questions. Cheers, -Benson On Mar 14, 2013, at 10:14, Robert Raszuk <[email protected]> wrote: > Hi, > > I think I am hearing quite clearly from most of the presentations that > there is sort of opinion that I2RS talks to RIBs from some form of > centralized controllers/route servers/topology collectors etc ... > > Does this mean that I can't use I2RS in completely distributed manner > without any central controller ? > > Note that more and more modern routers and switches gives customers > x86 local compute facilities on the box. So I can easily imagine that > I write a distributed application which communicates box to box at the > application level and installs locally via I2RS interface it's > "routes" to RIB. > > Is this model of operation out of scope of this WG ? > > Best regards, > R. > _______________________________________________ > i2rs mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/i2rs _______________________________________________ i2rs mailing list [email protected] https://www.ietf.org/mailman/listinfo/i2rs
