David Woodhouse wrote:
> On Fri, 2011-12-16 at 15:45 +0000, Andrew Benham wrote:
>> The router (running Linux) and another Linux box wouldn't route to it,
>> so on both of those I added a host route to 192.168.1.0/32 in addition
>> to the 192.168.1.0/24 network route already present.
>
> If you make it 192.168.1.0/23 instead, then packets to 192.168.1.0
> should work fine.
>
> For *real* addresses, that's slightly harder. For example if I wanted to
> use all four address of my 81.2.98.173/30 subnet, I could try bumping
> the netmask up to something like /24 and then doing proxy ARP tricks on
> the router so that the other addresses on that fake subnet, but not
> physically on my Ethernet, can be reached.
>
> However, that just moves the problem. I still wouldn't be able to reach
> 81.2.98.0 and 81.2.98.255, although I'm slightly less likely to want to.
>
> I suppose if I set the subnet mask to /0 and then do proxy ARP for the
> *whole* of the Legacy IP address space, that might work?
>
> Since I'm with A&A, the sanest plan is definitely just to ask for a
> bigger range and stop playing silly buggers 
>
Indeed. That was generally our conclusion when we need to implement such
policies.
And RIPE was sympathetic to the fact that we tended to eg,. put in a
proxy mail and web engine and firewall the uses on a private subnet . so
the machines we needed the numbers for were the outward side of the
proxies and relays, the router and the 'corporate web server'
Before NAT reared its ugly/elegant head depending on how you view it..
There is no doubt that without NAT we wouldn't have domestic broadband
as we know it and IPv6 would have had to be rushed in a lot quicker.
And its still a bi of an unkown donkey even now.