Hi, Robert, Thank you for the info.
Now I'd observe that everyone has different variations in the details of operation, although the basic idea of having IDR work on AS numbers. We might suggest everyone comes with his/her own draft based on this basic idea and compare the pros/cons of each proposals. The starting point might well be Bob Hinden's RFC1955 if people would agree. Different proposals would be weighed against routing scalability, mobility, multi-homing, easiness of deployment, etc. Maybe, Robin's criteria list might come into play again. Concern. Now that the group (chairs) have already chosen ILNP as the ultimate solution, I'm not sure if there's any value/use of exploring other lines of approaches. Perhaps, 'out of scope' or 'out of context'? Regards, DY On Sat, May 8, 2010 at 7:06 PM, Robert Raszuk <[email protected]> wrote: > Hi DY, > > Just for the record Enke Chen and myself have came up with this idea as well > about ... 4 years ago :) > > There are number of written documents we have on this. You can easily google > for it to see where is has been proposed before (even in this WG): > > http://www.ietf.org/mail-archive/web/idr/current/msg03174.html > > The main part of our proposal was that all transit ASes do not need to be > upgraded as IP header would contain plane old IPv4 address constructed from > AS number. Also all destination ASBRs would automatically recognize this > address and decapsulate. > > But our primary focus was not too address "melting Internet" ... our focus > was to provide a means and ability to provide fast connectivity restoration > aka PIC but in the inter-as scope (so far it works well intra-as). > > The mapping plane for it comes for free ... each BGP update carries the > origin AS today .. now we can even RPKI validate it - to eliminate manual > mistakes. Soon we will also extend current origin validation to make it even > more secure. > > Clearly the deployment price to pay would be to carry AS host routes oh well > ... perhaps additional 30K routes more in the Internet if everyone suddenly > would participate. > > But where and what is the benefit ...... The benefit would be only feasible > if for a given network all of it's customers and transits start sending AS > IP address encapsulated packets. Otherwise you still need to keep all > addresses in control plane as well as put all addresses in the FIB. > > But now that we have ILNP on the table I am just thinking loud here ... if > DNS record could return us not the locator or set of locators, but EID's > numerical representation of an AS number hosts which support it could easily > encap it this way while in the inner header still contain today's IP address > of the given server. > > So here is more of the question to ILNP authors if they have considered such > option .... > > Many thx, > R. > > >> Like... >> >> o Packets would be appended by an AS number of the destination at >> exiting its own AS. >> >> o DFZ routers sees only AS numbers in determining the next hop router. >> >> o The destination IP address would be examined only after the packet >> has entered the destination AS. >> >> Regards, >> DY >> >> >> >> On Sat, May 8, 2010 at 9:26 AM, Dae Young KIM<[email protected]> wrote: >>> >>> This sounds similar to an idea I raised some months ago; inter-domain >>> routing based on AS numbers. >>> >>> Some people were criticizing that AS's are not of equal scales (from a >>> tiny one to a global scale) and that it is not a useful/practical >>> granularity. >>> >>> Studying the IRON, the enterprise also can be as small as SOHO and as >>> big as the entire global Internet. Not much different. >>> >>> Then why not try an inter-domain routing (solely) based on AS numbers? >>> >>> Nonsense? >>> >>> Regards, >>> DY >>> >>> >>> >>> On Sat, May 8, 2010 at 6:13 AM, Toni Stoev<[email protected]> wrote: >>>> >>>> На Friday 07 May 2010 22:40:28 [email protected] написа: >>>>> >>>>> Toni, >>>>> I am sorry, but I don't understand what you mean. >>>>> >>>> >>>> Heiner, >>>> >>>> Currently inter-domain routing is based on IP address prefixes, which >>>> are intra-domain context, and on AS number paths. >>>> The simple and scalable model is the inter-domain routes to be only AS >>>> paths. >>>> >>>> Toni >>>> >>>>> >>>>> Heiner >>>>> >>>>> -----Ursprüngliche Mitteilung----- >>>>> Von: Toni Stoev<[email protected]> >>>>> An: IRTF >>>>> RRG<[email protected]>http://en.wikipedia.org/wiki/Classless_Inter-Domain_Routing >>>>> Verschickt: Fr., 7. Mai. 2010, 0:41 >>>>> Thema: Re: [rrg] RG futures >>>>> >>>>> >>>>> On Thursday 06 May 2010 12:02:39 [email protected] sent: >>>>>> >>>>>> How about an honest discussion that deals with the real causes of the >>>>> >>>>> scalability problem! >>>>>> >>>>>> The scalability (size) issue only unveils how miserably bad the (most >>>>> >>>>> important) IETF routing paradigms are. Particularly: >>>>> ... >>>>>> >>>>>> Heiner >>>>> >>>>> Intra-domain routing is as simple as node location naming follows >>>>> topology. >>>>> >>>>> Inter-domain routing has issues with intra-domain context – route >>>>> identification >>>>> prefixes based on intra-domain identity/location naming, so called >>>>> addresses. >>>>> Inter-domain routing has issues with route destination/endpoint >>>>> ambiguity – what >>>>> an inter-domain route leads to, node(s) or autonomous system? >>>>> >>>>> Regards to all. >>>> >>>> _______________________________________________ > > _______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
