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

David Arnold arnold at dstc.monash.edu.au
Wed Nov 21 21:25:04 EST 2001


-->"Darren" == darrend  <darrend at natwide.com.au> writes:

  Darren> Given the fact that an AP can only handle a limited number
  Darren> of connections, a class C allocation to each AP node should
  Darren> be sufficient.

an "AP node" corresponds to what the seattle people call a CxNode, if
i'm understanding you correctly.

   http://www.seattlewireless.net/index.cgi/CxNode

under the scheme i proposed, such a node would get a /24 (aka
C-class), as you suggest.  depending on the hardware (number of cards,
and hence number of channels, bridged onto the segment @ approx 20
clients/card max), such a node could handle more connections, but
there seems little point.

the other possible cause of address consumption is for clients using
more than a single address.  a /24 gives 32 /29s (each with 6 hosts),
for example.

so yes, i agree, a /24 for each AP node should be sufficient.

  Darren> Should they require more (ie nodes connecting to them wish
  Darren> to run smaller subnets), then just give them more. We can
  Darren> still use dns for identifying areas/AP nodes (see my
  Darren> previous post this thread), regardless of the IP's allocated
  Darren> to them.

we can (humans) but routing protocols can't, and that's what will be
problematic once we start linking the scattered clusters of nodes
together.

i'm really quite ambivalent about the DNS structure, since you can
have as many mappings as you like from DNS to an IP address.

the address allocation scheme is more concerned about being able to
route across a network of thousands of such semi-autonomous clusters.




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



More information about the Melbwireless mailing list