[e2e] architecturally speaking

Jon Crowcroft Jon.Crowcroft at cl.cam.ac.uk
Sat Mar 29 00:16:00 PST 2003


um, sorry, i wasnt clear

the delay challenged stuff IS more radical in a sense than plutarch is
as stated at the moment...i agree perhaps it isn't really orthogonal;
in that the context and intersticial function stuff we do is capable
of implementing solutions needed for DTNRG  but we hadnt got that far
out in our thinking....i guess its encouraging that there appear to be
several independent projects that are converging on similar things that look
like they could work and be built....



In missive <200303282339.h2SNdjSs023997 at dwight.CS.Berkeley.EDU>, Kevin Fall typ
ed:

 >>> plutarch is not about naming and addressing much - if you want a good
 >>> solutuon to that problem i would refer you to paul francis work
 >>> (either his thesis or more recent work or both) - it is more about
 >>> what it says so please read it:-)
 >>
 >>>From what I can tell the case being made is that the current Internet architecture
 >>has problems when faced with 'radical' heterogeneity, and so you group homogeneous
 >>portions into contexts and you use some form of functions
 >>(IFs) [mappings in the case of naming/addressing,
 >>possibly maintained by soft-state and registrations, etc] to provide the
 >>bridge between contexts.  Section 4 (the section titled 'Architecture')
 >>discusses naming, addresses, and contexts.  So if this architecture isn't about
 >>naming and addressing much, could you please elaborate?
 >>
 >>> ...
 >>> the DTNRG stuff , which we read is about disconnection.
 >>
 >>Well, while somewhat true, a bunch of time has been put into thinking about
 >>'radical' heterogeneity (I do like this term, btw), and we would hope
 >>that things work reasonably well over not-so-disconnected and not-so-very-high delay
 >>environments.
 >>
 >>> ...
 >>> overlays are very f ine - overlays for disconnection have to do state
 >>> (context) management very well - we skate aroud that coz we only need
 >>> to do it "on the order" of as well as the internet or so...
 >>
 >>So, if you are 'on the order of the internet or so', perhaps it isn't
 >>so radically heterogeneous ;-)
 >>
 >>Anyway, the reasons for the dtn overlay are two-fold:  one for heterogeneous
 >>operation over very different kinds of networks and two for dealing
 >>with disconnections and interruptions.
 >>
 >>> 
 >>> 
 >>> good luck - your parsec-age may vary...
 >>> 
 >>> btwe,m i really like the DTNRG stuff , but it is orthogonal
 >>
 >>thanks, but I believe many of the issues are essentially the same, so I don't see
 >>them as particularly orthogonal.
 >>
 >>cheers,
 >>- K
 >>

 cheers

   jon




More information about the end2end-interest mailing list