Re: router-router NHRP

[Dimitry Haskin <[e-mail]>]

Yakov,

> 
> > > >   Another case which I believe is not supported by the proposal is
> > > >   "third party advertisements", i.e. when an exit forwarder is not
> > > >   the same device as a router that acquires a route.  But may be it is
> > > >   not an important case to support.
> > > 
> > > Actually the proposal does support this case. So, if a border router
> > > receives a BGP route (which carries NEXT_HOP information), the border
> > > router may pass the next hop information (rather than its own address)
> > > in response to an NHRP Request.
> > > 
> > 
> > But, in this case, how would forwarders at each end of a shortcut monitor
> > the viability of the route that was used for the shortcut?  Since they don't
> > run BGP and only have the default route in their FIBs they wouldn't see any
> routing change unless they are somehow informed of such a change by the border
> > BGP router which responded to the Request.  The shortcut's egress forwarder
> > would't be even able to associate a shortcat with a specific route since it h
> as
> > never seen the corresponding Request in the first place.  Isn't it so?  I did
> n't
> > see any mechanism in your proposal to handle that.
> 
> Consider a router X within an OSPF domain, ASBR Y of that domain, and
> router Z (outside of the OSPF domain) that peers with Y using BGP.  Z
> advertises to Y a route to some destination D, and puts its own address
> as the next hop. Y injects this route into OSPF.  When X originates an
> NHRP Request for D, and the Request reaches Y, Y has a route to D, but
> the next hop on this route is not Y, but Z. Y returns Z as the next hop
> in the NHRP Reply. Since Y certainly sees any changes in this route, Y
> would be able to send to X a Purge message once Y notices the changes.
> 

Good.  But there is no text in your proposal that requires a proxy router that
replied with a third party address to send a Purge message to the originator
of the shortcut.  The proposal should be clear on that since there are
certain implications of such a requirement, e.g. the amount of state that need
to be saved at a proxy router, especially if such a router is a proxy for
a good number of edge forwarders.

Dimitry