[e2e] Lost Layer?

[Jon Crowcroft]

agree - in fact, the model of layers is obsolete in terms of s/w APIs for a
variety of reasons - when we did Haggle, we had a set of cooperating agents
to build up a protocol as needed - others (handley's protocol heaps, for
example) have looked at this - also, in terms of interfaces, one can use deep
reflection to modify internals of a protocol through an API, rather than
being restricted to modifying things above a restricted service interface...
so any OO trained programmer ought to be fine with that idea..

finally, a lot of cooler things can be done with typesafe systems in terms of
puttign new code into a protocol "stack" (or heap) safely with contracts
and/or proof carrying code - or else just build a whole lot of compeltely
different comms systems as needed out of lots of micro-protocol
components....(see
http://cacm.acm.org/magazines/2014/1/170866-unikernels/abstract
for this approach for cloud)

so yes, we have much better tools for extensibility nowadays...and could go a
lot further in future proofing things

In missive <20140111.143902.71090496.sthaug at nethelp.no>, sthaug at nethelp.no typed:

 >>> > you need a sublayer convergence (as per day's work) but also the
 >>> > socket layer needs revising badly to allow for a wider set of
 >>> > transport service semantics than came out of the fast 
 >>> > hack that bbn and berkeley did
 >>> 
 >>> Shouldn't we agree upon a model and then upon offered services and APIs
 >>> first?
 >>
 >>Maybe not, because you can never foresee all of the services and APIs
 >>you might need in the future. You need an extensible model.
 >>
 >>Steinar Haug, Nethelp consulting, sthaug at nethelp.no

 cheers

   jon