[MLB-WIRELESS] ip allocation list based on suburb

[Roger Venning]

Dave Keller wrote:

>Why not just allocate the ips whichever way required and use reverse DNS to
>give more meaningful geographic information?
>
<snip>
How about we use a real address allocation tracking mechanism like 
whois? e.g. like http://www.apnic.net/apnic-bin/whois2.pl

On the wider issue of address allocation it turned out I had a bit to 
say so quick summary:

I don't think aggregatable addressing will successful on any finer 
granularity than Melbourne as one aggregatable block, Geelong as 
another, etc. We should just give each Melbourne node a 172.20.0.0/20 
address, given out sequentially as a node goes to 'testing' status. Also 
sequentially allocate Melbourne nodes subnets from the 10.1.0.0/16 as 
requested. Aggregatable address allocation will prove impossible in a 
mesh network.

Long:
There are several factors at play in IP address resolution

- each host on the WAN must have its own unique address within the WAN
- it is desirable to also have uniqueness across other networks that may 
be later interconnected, e.g. other wireless community WANs. Also the 
Internet, but in this case it is essentially impossible due to the 
restrictions around obtaining large IPv4 address blocks.
- it is desirable (and on some scales *required*) for addresses to be 
aggregatable. However I honestly do not believe that this will be 
possible in a mesh style network on any basis other than for instance on 
the interfaces between say Melbourne and Geelong, etc. There will 
(hopefully!) be too many diverse paths for aggregation to occur within a 
heavily 'cross-linked' network that will not have obvious levels of 
hierarchy.
- hierarchical address allocation assists distributed administration of 
handing out addresses. However I think it is tenable to put together an 
CGI style request system that takes details (e.g. email address, node 
ID, desired size of subnet (to support wired and wireless 'clients' on a 
'home' network surrounding the node) that deals with all requests for 
say Melbourne - so no council based system is useful.

Can somebody prove to me that is possible to have hierarchical / 
aggregatable addressing without requiring a corresponding strict network 
hierarchy (ie. nodes on bottom layer only connect to next highest layer, 
not each other, next highest layer connects to its superior layer in a 
similar fashion, ad naseum)?

I think that we should just do this:
- set up the node maintenance page so that when the owner takes their 
node to 'testing' status they are allocated a node IP per wireless 
interface and asked if they also want a network block (with provision to 
later request further network blocks (more local clients) and node IPs 
(further wireless interfaces)
- the node IP comes from the 172.16.0.0/12 (RFC 1918) space. This could 
simply be numbered sequentially by non-contiguous urban area, rural 
areas handled all from the one pool.
    - 172.20.0.0/20 Melbourne
    - 172.20.16.0/20 Geelong
    - 172.20.32.0/20 Rural Vic (I know there are also regional centres, 
can be handled with new blocks on an as needs basis. Hell I even come 
from one - or maybe Hamilton is not big enough to be a regional centre?!)
    - the node addresses are assigned to the wireless interfaces. If a 
connected subnet must be used, then it is the whole 172.20.0.0/20 for 
Melbourne for instance. A dynamic routing protocol such as OSPF is used 
to establish host routes in the routing table to each and every node 
within this autonomous system of the WAN
- network blocks taken from 10.1.0.0/16 for Melbourne and simply handed 
out sequentially as requested in blocks, e.g. 10.1.0.0/28 first, then 
10.1.0.32/27 (must start on subnet boundary), then 10.1.0.16/28 (taken 
out of gap)... etc.
    - these network blocks are advertised through dynamic routing 
protocol into the WAN. Only given out to clients that do not route on 
the WAN (both wired & wireless non-router clients).

Roger.

-- 
-------------------------------------------------------------
Roger Venning	\ Do not go gentle into that good night
Melbourne        \ Rage, rage against the dying of the light.
Australia <r.venning at bipond.com>                 Dylan Thomas




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