Re: [RAM] Comment on draft-farinacci-lisp-00.txt (LISP)

[Eliot Lear <lear at cisco.com>]

Yakov,
> What you called "end-end names" are not really "end-end names" but
> locators. It just the scope of these locators is not the whole
> Internet. What you do at the boundaries is mapping of site-wide
> locators into the ISP locators. 
>   

I think this is a great way to look at the engineering problems we are 
targeting.  Scope in the case of LISP indicates the appropriate lookup 
mechanism.  How that scope is determined is something of an open 
question.  For instance, one could say that an address falls within the 
global routing scope if an entry for it exists within the RIB.  One 
could say that an address falls within the global LISP scope (for lack 
of better words) if an EID/RLOC mapping can somehow be determined.

The benefit of such an approach is that it keeps us out of all the 
pitfalls that application developers have warned us for years.  We 
retain transparency, retain the same APIs, and we provide for a 
transition.  Here is the real question: how responsive can the network 
be in the face of a failure, and what is the restoration time from that 
failure?  Therein lies the rub.

By the way, I still believe we should invest more research efforts in 
HIP, and API issues relating to a LOC/ID split that occurs on the host.  
A lot of the standardization would not normally happen under the 
auspices of the IETF, but rather in the IEEE under POSIX.  It'd be sort 
of funny if we made this their problem ;-)  And then funnier when a 
bunch of the same faces showed up there, making it our problem again :-)

Eliot

_______________________________________________
RAM mailing list
RAM at iab.org
https://www1.ietf.org/mailman/listinfo/ram