[e2e] DNS: Directory or Rendevous?

Jon Crowcroft Jon.Crowcroft at cl.cam.ac.uk
Wed May 19 01:45:04 PDT 2004


so another idea regarding the rendezvous and name space (and possibly nat
traversal)


 >>In message <0c3501c42491$e8dde2d0$333924a0 at laptop60>, "Micah Beck" writes:


...so i shouldnt say this but - like 
"fire, women and danerous things"
(google for that - you will be amazed)
this idea is  like the "come from" statement of routing:

loose destination routing...
name is what you are 
is bnound to
address which is where you are
is looked up to find a
route, which is how to get there

so a route is not necessarily exact or uniq - 

loose destination routing is similar in concept to dns rotaries (lookup a name and get different answers determined
by domain name service authority based on who asks)

this is same idea as mothy roscoe's capabilities, but lighter weight...
and harder to beat with collusion/replay attacks...

so implemtation is easy - to do loose source routing, end systems take part 
in routeing algorithm in the sense that
they take part in receiveing link state (or path vector) updates - 
but instead of being at the mercy of the
router/isp consipracy to figure things out, a sender/source compoutes a route out of possible choices

one implentation idea:
in loose destination, a receiver computes a route but only advertises part of the route in the clear, and
obfuscates the rest (like onion routing) and leaves state "out there"
(breadcrumbs, or transient landmarks) for routers to decrypt 
and forward

in another implemation, we might imaging a Realm Specific IP system 
or collection of NAT traversal hacks, implemewnationg the same thing...

sort of also like the internet indirection stuff...

cheers
jon 


More information about the end2end-interest mailing list