On Mon, Nov 12, 2012 at 1:45 AM, Danny McPherson <da...@tcb.net> wrote: > > On Nov 10, 2012, at 12:54 PM, Tony Li wrote: > >> >> >> I agree, but as you correctly point out, SIDR is an engineering solution. >> If you dislike that particular solution, you're of course free to propose >> others. However, the correct forum for engineering solutions is the IETF. > > I was hoping for some research and fresh thinking related to > rate+state+autonomous operations &security in inter-domain routing, hence my > message here. I don't have a solution envisaged, I just know what I'm seeing > in current work is err.., going to present some challenges. > > I was hoping for "architectural alternatives, research and experimentation in > secure routing architectures and algorithms being encouraged, with an aim to > understand whether new directions can provide effective solutions, to work > out candidate designs as necessary for a complete solution, and to fully > understand both the gains and the tradeoffs that the new solutions may > bring." (some of that text may sound familiar :-) > > As a side benefit, whether the routed protocol changes or not, using security > as the fulcrum may well help foster broader consideration of alternative > routed protocols... > > Else -- some work into scalable *static* inter-domain routing techniques and > solutions might be a worthwhile endeavor *8^/ > > If that's not within RRG scope then I suppose I understand how we arrived > where we currently are... > > -danny >
Hi Danny, if some of the RRG proposals (LISP, ILNP, IRON, hIPV4) gets globally implemented then the endpoint locators (EID, ELOC etc) will not show up in the DFZ and the global routing table should be more static than today. However, none of the RRG proposals have a proper solution for inter-domain traffic engineering, thus I think your concerns are right on spot. It would be interesting to take the IRS approach (as an interface to the FIB) and design from scratch an inter-domain traffic engineering architecture, where the ISPs can create explicit paths for certain traffic flows/bundles between domains without inserting more granular prefixes in the RIB of DFZ. This new inter-domain architecture would be designed as an additional control plane in parallel with the current control plane of the existing BGP based DFZ. Is this doable, what are the pros&cons? Patrick _______________________________________________ rrg mailing list rrg@irtf.org http://www.irtf.org/mailman/listinfo/rrg