In einer eMail vom 24.02.2009 22:05:43 Westeuropäische Normalzeit schreibt [email protected]:
As always, other insights into the problem, discussions of architectural conclusions that we can clearly draw and other conceptual items leading to an architectural proposal are most welcome. There's a lot more than could be in this document outside of just taxonomy (e.g., mental question du jour: do identifiers need to be globally unique?). Earlier I gave two answers to above "question du jour" wrt strategy C: While strategy C (TARA) is not yet completely deployed, IPv4 addresses have to be globally unique, but not thereafter. Globally uniqueness of IPv4 addresses is definitively needed for the incremental deployment phase. The reason: The hereby employed router's view of a well reduced/sparsed internet topology will also consist of non-TARA GRE-tunnels which interconnect isolated TARA subnetworks. The computation of these GRE-tunnels rely on the global uniqueness of IPv4. Having said this, I like to ask the following question du jour: Does jumping ahead with some particular strategy (e.g. with LISP/ALT) harm all the other potential solutions ? Speaking in favor of strategy C, the answer is " Yes, it harms". Its objective, to reduce the classic BGP tables continuously, means: From some point in time a TARA-router should stop disseminating its own user reachabilities. Concurrently any other TARA-router will have to do what is called default-mapping in ALT: it has to advertise reachability with prefix length 0 as to attract IP flows from the non-TARA surrounding. Hence, there is a competition with ALT-routers who do the same. Of course, by chance,a TARA-router may be in the way to some ALT-router -- and vice versus. Because I think that ANY strategy needs a concept for incremental deployability, i.e. might need that default-mapping mechanism. It is indeed a pretty general and not a ALT-specific mechanism. IMHO, this aspect should be considered prior voting for a LISP working group in the IETF. So I repeat my question du jour: does LISP/ALT jumping ahead hurt ? Heiner
_______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
