Re: [RAM] The mapping problem: rendezvous points?

[[e-mail]]

David,

David Conrad <[e-mail]> wrote on 10/05/2007 00:20:56:
> 
> People have been talking about having a push based mapping 
> mechanism.  They have also been talking about having host-level 
> granularity.  How is this going to work?  Are people imagining a 
> flood protocol that propagates host-level mappings to every host on 
> the Internet?

I am the proponent of a BGP-based push mapping mechanism. In my
proposal on TIDR, I mention the possibility of propagating FEC
mappings that could even descend down to the TCP or UDP port level
per host. Obviously this is a nonsense nowadays. But, as almost
always, not all that can be done is eventually done.

I wanted to reflect in my proposal all the potentials of TIDR.
But this is similar to what currently happens with the use of
BGP for routing in the Internet: the BGP specification says
nothing about the maximum mask length that should be stored in
the RIB. It is the best current practice which dictates for
example that /24 is a reasonable limit. And this is achieved
by using filters.

Therefore an important feature of a flooding mechanism should
be the ability to differentiate the relative importance of
the objects being mapped so as to treat them differently by
means of filters of some type.

Someone mentioned some days ago that host-based solutions and
network-based solutions are orthogonal. I think he's right.

Regards,
Juanjo

_______________________________________________
RAM mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/ram