Re: [Idr] ietf-104 responding to Rüdiger's comment about bfd strict mode

[Robert Raszuk <[e-mail]>]

Hi Jeff,

Yes clarification of the scope is needed.

But my broader point also was to clarify which operational model fits best
to different scenarios.

We have clearny two options:

* keeping up/down of bgp session on say loosy links

vs

* extend the notion of valid/invalid next hop in the same env.

The latter seems much more universal and applies to various deployment
models.

Former can be used for specific use cases ex: directly connected ebgp peers
or in more general it can be used where bgp control plane is congruent with
data plane. That often for ibgp is not the case.

Thx,
R.



On Thu, Mar 28, 2019, 16:20 Jeffrey Haas <[e-mail]> wrote:

> Robert,
>
> On Thu, Mar 28, 2019 at 03:26:09PM +0100, Robert Raszuk wrote:
> > Yes you have bfd for bgp. Use case is for ebgp to bring bgp down as soon
> as
> > your peer goes down without cranking up bgp keepalives.
> >
> > Well as it was presented today the scope of current proposal is much
> > broader. Hence the comments.
>
> I find it likely that the draft needs some clarity as to whether we'd want
> this capability present (or not) when BFD is not intended to be used on the
> link.
>
> Your example of "we're doing ibgp, please don't use BFD for BGP":
> - If we did want to use BFD, please use strict.
> - If we didn't want to use BFD, we might still want to specify strict as a
>   capability so that by not having it come up you didn't end up in one of
>   your interop scenarios.
>
> -- Jeff
>