[frr] [PATCH 01/11] bgpd: BGP VRF processing handling

[Philippe Guibert]

Hi,

On Mon, Feb 6, 2017 at 8:27 PM, Jeff Tantsura <jefftant at gmail.com> wrote:

> RD is usually configured in context of VRF, not per prefix. What would be the use case?

I agree with that. one RD per VRF.

> The best way to address prefixes is to do so indirectly, thru route-map/prefix-list/your preferred-feature-name

Using route-maps or prefix-list are good tools to apply some behaviour
on incoming our outgoing NLRI messages.

> prefix/label/rd combination makes little sense to me, label per prefix is not really good thing unless there’s a need (we could go in details when), default mode - label per NH, perhaps with knobs per VRF (would require additional lookup if there are more than 1 NH’s), per-prefix, RD per prefix - see comment above

Usage of route-maps has been proposed, and some more specification is
not yet achieved.
For more information, see:

https://docs.google.com/document/d/1w_ie2tNXCgn0N3ZNFGYTK6lJkwMmk_XN5yz33MMNNqM/edit#

> How would additional ext-comm added, think of RO/SOO -> having policies within import/export stanzas my be desirable

Currently, there is a command available under route-map:

bgpd(config-route-map)# set extcommunity soo
  ASN:nn_or_IP-address:nn  VPN extended community


> Type 5 - I like Junos’s keyword - "ip-prefix-routes” with associated attributes
>
> my 0.2c
>
> Cheers,
> Jeff

Regards,
Philippe