Re: [i2rs] RIB Model draft

Tom,

A discussion would be great!  Why don't you start it?

Alia

On Fri, Oct 11, 2013 at 4:57 AM, t.petch <ietfc@btconnect.com> wrote:

>
> ----- Original Message -----
> From: "Nitin Bahadur" <nitinb@juniper.net>
> To: "Joel M. Halpern" <jmh@joelhalpern.com>; "Russ White" <russw@riw.us>
> Cc: <i2rs@ietf.org>
> Sent: Thursday, October 10, 2013 7:13 PM
> <snip>
> >
> >AS Paths are a BGP construct.  If you are not running BGP then the
> >question above (about two paths from the I2RS agent) would seem to
> >apply.  If you do have a BGP running, then the BGP is usually where
> that
> >sort of policy lives.  My concern is that this mechanism assumes a RIB
> >agent that handles BGP specific operations.  I would prefer to leave
> BGP
> >mediation to the BGP model.
>
> There were folks who had expressed interest in inter-domain paths.
> That's
> why this
> was added. I agree with you that the handling logic will be BGP'ish and
> there is an
> implicit assumption that the RIB agent can handle this.
>
> However, it's obvious (at least to me), that most of the use-cases for
> I2RS will be
> intra-domain and a device that does not support inter-as/bgp should be
> fine for
> most of the deployment use-cases.
>
> <tp>
> I keep on thinking that producing designs without agreed requirements,
> use cases, is often a costly mistake.
>
> My thinking on use cases was quite different, but unless and until we
> have the discussion on eg
>  draft-white-irs-use-case-00.txt et al.
> who knows?
>
> Tom Petch
>
>
>
>
> <snip>
>
>
> _______________________________________________
> i2rs mailing list
> i2rs@ietf.org
> https://www.ietf.org/mailman/listinfo/i2rs
>