> Typically, the answer is 'not really'. DHCP isn't really well-suited
> for ad-hoc wireless networks.
The MITRE people seem to have addressed that much better as far as I've
been able to tell (difficult to tell at 2 in the morning on a bear of
> That's a different layer; IMO, now that we have a
> sufficiently-advanced layer 2, we can build the higher layers on that
> and not have to change them much when we get a better bottom layer.
I think the layer abstraction approach can only get you so far. You can't
be totally hardware-agnostic if you want to go places. Not on the long
It's clear that our universe imposes constraints like relativistic latency,
degradation of link quality over distance and hence predominance of short
multi-hop paths, which requires design of the protocols so that you can
get cut-through routing over meaningful distances, and are also able to
ping a machine in .kuiper or .oort. So you have to use precomputed paths,
where headers directly switch the selection in the crossbars while being
consumed in the process, or stuff which caches the header in a short
FIFO (a short loop of fiber as delay line, in purely optically switched
network), while the router decides which space radiant to cut the message through.
> This is the first really on-topic discussion on FoRK in a month, and
> you want to take it off-list?!
I agree, but the other folks don't seem to be too enthusiastic.
I'd wish I hadn't a job to attend to and a flat to renovate, this matter
warrants at least few weeks of undivided attention. We're late, but it's
probably still not hopeless.
Btw, the IPv6 peope boasted how many nodes/m^2 of Earth surface they'll be
able to address, but this doesn't involve um^3 nodes in a sphere lighthours
Still, any clue as how to intelligently map a WGS84, including altitude into
a 128 bit IPv6 address? 32 bit/degree of freedom, it does roughly give you
1 mm resoultion for Earth surface, right? That might be enough.
This archive was generated by hypermail 2b29 : Fri Apr 27 2001 - 23:14:02 PDT