[e2e] end2end-interest Digest, s/Vol 53, Issue 10 / sacred cows, sacre bleu/

[Jon Crowcroft]

yes, sometimes you need some mapping and indirection - that isn't terribly
surprising since
the purpose of this proposal is to do parsimonious
ipv4 support for mobility, multihoming and
multipath - these typically all need some extra information somewhere (even when
you add route optimisation in mobile ipv6) - the cellular guys have this all
sorted but ip folks dont seem to be willing to bite the bullet on the necessary

actually it turns out in some cases (actually quite a lot) we dont need
an explicit indirection service - it can be implicit in how packets are routed
(ie. the snark has a nat like function in some cases)

thanks to everyone for all the input  = i will now try and clarify al the points
made in an update to the paper in the same place

In missive <4877AC87.5060104 at employees.org>, Scott Brim typed:

 >>On 7/11/08 10:55 AM, Jon Crowcroft allegedly wrote:
 >>> by the way, i dont put an _address_ in the transport - i put an identifier - this
 >>> is then used receivers to lookup an address to the originator. i _might_ put a
 >>> routing hint (e.g. a care of adddress) in the transport too - this is  
 >>> not the same as putting in an actual address (which i am getting rid of as it is
 >>> invalid MOST of the time...
 >>
 >>That's what I thought and what I've been wondering about.  The recipient 
 >>of the identifier needs to look it up somewhere.  99% of the endpoints 
 >>out there today don't have global names.  Where will the identifiers be 
 >>looked up?  Either every endpoint gets registered in something like DNS 
 >>(and it's updated dynamically) or you depend on application-level 
 >>intermediaries.  Either way you're depending on intermediaries and 
 >>you've lost e2e transparency?
 >>
 >>Scott
 >>

 cheers

   jon