[e2e] Overly Overlay; Peer to peer is commonplace

[Christian Huitema]

>  >>> From my perspective, what's interesting is that it looks like the
>  >>> mobile ad-hoc networking folks are getting ahead of the wire-line
>  >>folks.
>  >>
>  >>Uh, looking at recent publications, it seems that most ad hoc
routing
>  >>algorithms scale with the number of nodes N in at least O(N), if
not
>  >>O(N^2) -- they also have a scaling dependency on the number of
> neighbors
>  >>per node. Not quite what you would want for Internet wide
routing...
> 
> not sure what publications or what definition of "recent" you are
> using - i've seen self organisign hierarchies (not just spine) based
> ad hoc protocols which do way way better....depends on the _mobility_
> model - a lot of people have naive mobility models - if you use a mix
> of techniques, and a mix orf appropriate scale mobile/wireless node
> velocity models, you can do really quite well...

I was looking at the MANET drafts. I can see two kinds of protocols,
either based on pre-computation of routing tables with one entry per
destination, or "flood and trace" variations. The amount of state per
node in the pre-computation variant scales as O(N), and so does the
number of message per node; the total number of messages in the network
scales as O(N^2). In the flood and trace variant, the cost of a given
flood scales as O(N) and the number of floods scales as O(P), where P is
the number of peers maintained by a given host; you can argue that P
scales as O(N), or possibly slightly better (but not O(1)!); in any
case, the number of messages in the network grows faster than the number
of nodes. This may be perfectly fine for the ad hoc networking
application, but I would not take it as a model of the general solution.

-- Christian Huitema