We got it.

On Wed, 19 Apr 2017 at 20:45, Alvaro Retana (aretana) <[email protected]>
wrote:

> On 4/19/17, 10:03 AM, "Jared Mauch" <[email protected]> wrote:
>
> Jared:
>
> > Are you saying that IOS-XR is non-compliant with 9.1.1 because it does
> not have “bgp
> > unsafe-ebgp-policy” as the default?
>
> No.  I wasn’t talking about any specific implementation.
>
> One of the reasons I like your document is the fact that rfc4271 is not
> specific about what should be done if there is no policy – it just says
> that policy may be applied.  From that point of view, the XR implementation
> is neither compliant nor non-compliant.
>
>
> > At what point does the Cisco implementation make that decision?
> >
> > We seem to be triangulating on where in the exact decision process
> people are
> > considering a route feasible or ineligible, can you speak to your
> implementation?  That
> > may provide guidance in documenting the IOS-XR practice.
>
> The XR implementation makes the decision while processing the Update
> messages.  Specifically, it drops all the routes from a peer without an
> incoming policy configured.  IOW, it acts as if a deny-all filter existed.
>
> In the conceptual model in rfc4271, the routes are not even put in the
> Adj-RIB-In.
>
>
>
> Compared to the current text (in -05), there would be no routes to mark as
> ineligible in 9.1.1.   The text I proposed also assumed that the routes
> would at least be received/stored.
>
> If you want the same behavior as XR then you would have to make the change
> in Section 9. (UPDATE Message Handling).
>
> Alvaro.
>
> _______________________________________________
> GROW mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/grow
>
_______________________________________________
GROW mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/grow

Reply via email to