[MLB-WIRELESS] [proposal] IP allocation (was: DNS, DHCP, static ip... )

[David Arnold]

-->"Drew" == Drew  <drew at no6.com> writes:

  Drew> wouldn't there be problems on every edge of each "geographical
  Drew> area"? if a node is outside your geographical area, but is
  Drew> actually the closest node to you, and you're not routing that
  Drew> way, won't that create problems/inefficiencies? 

yes.  typically, you obtain your address range from your upstream
provider.  in our case, we have no "providers" yet (ie. no backbone
links) and so this doesn't work.

address ranges that are connected outside their enclosing range (and
multi-homing is the normal culprit here), need to expose routes to the
whole world.  ie.  if my place is 10.10.10.0/24, and i connect to the
owner of 10.10.0.0/16, then the rest of the world only needs to know
how to reach 10.10.0.0/16 and from there the packets find their way to
me.  this means only one routing table entry for me and whoever else
is connected via 10.10.0.0/16.

small routing tables make for cheaper router hardware ;-)

the idea of rough geographical areas is that nearby people, who are
likely to connect, are also likely to share a long-ish portion of
their prefix.

but there will always be boundary cases which will fatten up the
routing tables.

as i said in the proposal -- maybe we don't have to worry?  if we only
expect to grow to size X, and BGP/OSPF/IS-IS/whatever can happily deal
with routing tables of size X without requiring gigabytes of RAM and
CPU in the routers, then all this becomes moot.

i don't have the operational experience to answer that, but i was
hoping someone here would ;-)


  Drew> seems to me having everyone run something like mobile mesh
  Drew> that automatically handles all of the routing would be the
  Drew> ideal solution.

i haven't yet read through the drafts, so i cannot comment with any
authority on its applicability yet.

in the overview pages i skimmed, it seemed that it was aimed at
*mobile* nodes, and i wonder if that means it is less suitable for our
usage in which most of the nodes will be stationary?





d

--
To unsubscribe, send mail to minordomo at melbwireless.dyndns.org with a subject of 'unsubscribe melbwireless'  
Archive at: http://melbwireless.dyndns.org/cgi-bin/minorweb.pl?A=LIST&L=melbwireless
IRC at: au.austnet.org #melb-wireless