----- Original Message -----
From: "Jeffrey Haas" <[email protected]>
To: "t.petch" <[email protected]>
Cc: "Russ White" <[email protected]>; "'Susan Hares'" <[email protected]>;
<[email protected]>; "'Jeffrey Haas'" <[email protected]>; "'Edward Crabbe'"
<[email protected]>
Sent: Tuesday, June 17, 2014 2:15 PM
Subject: Re: tables was: Re: [i2rs] draft-white-i2rs-use-case-05.txt has
been posted


> On Tue, Jun 17, 2014 at 10:11:01AM +0100, t.petch wrote:
> > It is B that I am doubtful about. It is fine for BGP or RIP, but I
do
> > not see it for link state protocols, that is for me, the SPT is not
a
> > table of routes, backup routes and so on but, well shortest paths.
> >
> > Do you see an IS-IS, or OSPF, table comparable to the BGP table?
>
> Without making specific comment on what *should* be present in the
model,
> there are three classes of clearly (IMO) useful information available
from
> the IGPs:
>
> - The LSDB, which provides topology
> - The TEDB
> - The active route for a given destination at a given node as per SPF
>   computations.
>
> The third item is RIB-like and is the usual input to broader router
> route-selection of an active path from multiple candidates.
>
> The third item will probably have tie-in to our RIB models.
>
> The first two items are places where abstractions in the model may
make
> sense, but there's also some benefit to simply exposing protocol
mechanics
> in the models.  The difference is "this is a type-X LSA in OSPF" vs.
"this
> is a link, this is a node, here are how they connect".
>
> The abstraction is a bit more useful in an I2RS context.
> The protocol mechanics are likely to be something that gets a protocol
> specific module in the owning Working Group.
>
> There will be interesting overlaps in the above two, and one of the
more
> interesting bits of coordination work I2RS will have.

Jeff

We have just got a netmod-isis draft
draft-litkowski-netmod-isis-cfg-00.txt
and I look for items of the third type, and at first sight, I do not see
them, just as I cannot recall seeing them in SMI models, either
standards-based or proprietary (unlike the LSDB).  Of course
they can be added but ....
well, I have said enough about it.

Tom Petch






>
> -- Jeff

_______________________________________________
i2rs mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/i2rs

Reply via email to