Re: [homenet] draft-boutier-homenet-source-specific-routing-00

[Henning Rogge <[e-mail]>]

On Wed, Jul 10, 2013 at 5:44 PM, Juliusz Chroboczek
<[e-mail]> wrote:
>> I am looking at the problem from the link-state point of view.
>
> You're assuming that non-source-specific routers drop source-specific
> routes, right?

Doesn't really matter, at least for the routes to the gateways.

>> Lets assume there is a random mixture of source-routing aware and
>> non-aware routers. The only restriction is that there is a
>> source-aware path of routers between all gateways (gateways to the are
>> of course source-aware too).
>
> Yeah, that's exactly what we describe in 4.1.

Okay... wasn't sure about it.

>> Every non-source-aware default route will bring the traffic into the
>> source-aware region around the gateways, which mean they should reach
>> the desired gateway (even if it hit a different gateway first).
>
> Yes.  Hence the language at the end of 4.1.
>
> A related question is whether it is okay to reannounce a source-
> specific route as a non-specific one, for the sake of legacy routers.
> (This is analogous to aggregation -- reannouncing a more specific
> route as a less specific one.)  The general answer is no, so I was
> wondering if you'd identified a class of topologies where it is okay
> to do so.  Does anyone know any good papers describing when aggregation
> is safe?

In my example I assumed that every gateway announces a
non-source-specific route too, so there should be no special code
necessary for the "normal" routers.

In terms of OLSRv2 an IPv6 gateway could announce its ::/0 route and
add a "src prefix xyz::/64" attribute (TLV) to it. Routers that do not
understand this attribute just ignore it.

Henning Rogge

-- 
We began as wanderers, and we are wanderers still. We have lingered
long enough on the shores of the cosmic ocean. We are ready at last to
set sail for the stars - Carl Sagan