On Tue, May 20, 2014 at 09:23:13AM -0700, Sriganesh Kini wrote: > On Tue, May 20, 2014 at 8:06 AM, Dean Bogdanovic <[email protected]> wrote: > > > Changes done by I2RS agent should have only device impact, not network > >> impact. If there is desire to change across network, then each I2RS agent > >> on each device should be updated. > >> > > > > It is not simple to classify a change to a device as having an impact > > scoped to just that device. E.g a route that is added to a device with a > > different preference can have network-wide traffic impact. > > > > > > How can I explain this better. The static route on device will be routed > > to the next destination, but at the next destination the dynamic protocol > > will take effect, unless, another static route directs the traffic > > differently. Makes sense? > > > I think what you really wanted to say is - An I2RS agent must not interface > with an I2RS agent on another network element. If an I2RS client requires > to interface with I2RS agents on multiple network elements, it must do so > itself without putting requirements on the I2RS agent on one network > element to interface with the I2RS agent on another network element.
As mentioned elsewhere, I tend to perceive I2RS's impact on the network as being similar to that of a routing protocol. If I want something I do to a given device to propagate, the state that is injected must either be suitable for injection into another routing protocol (redistribute into BGP, or IGP, e.g.) or must be fully local. We will have cases for both. For Sri's point, if I inject state in to router A and it overlaps with state in router B, it better be because that state had arrived via a routing protocol. In such a case, the interaction between I2RS and the protocol would already be defined. -- Jeff _______________________________________________ i2rs mailing list [email protected] https://www.ietf.org/mailman/listinfo/i2rs
