[c-nsp] using the first and last ip address of a range > /24 in a local pool

[Christopher E. Brown]

On 7/25/10 6:12 PM, Brett Looney wrote:
>>> Try connecting from a XP workstation to a .255 target address
>>> that is on a class "C" address. It will fail every time
> 
>> Wrong, running XP SP2
> 
>> D:\Tracking>tracert 192.0.2.255
> 
> Traceroute is not actually connecting. Try getting IE (or any other browser)
> to connect to a host with an address like that. The packets never leave the
> source. We had to change a customer implementation because Macs and other
> sensible OS could connect but Windows XP hosts (many of them - all running
> SP3) could not.
> 
> And if you read the Microsoft technote about it, they will tell you it
> doesn't work for everything XP SP3 and earlier and it is fixed in Vista.
> 
> B. 


I has been a while, but 98 through at least SE2 was clean for 0 and 255
where they were in the middle of a network.


The historic issue was cisco devices post ping flood plauge.


It is sane for the router handling the last hop delivery of a subnet to
decide what to do about network and broadcast adress (has an interface
in the network).


It is not sane for a middle router without any interfaces in the dest to
decide to treat 0 and 255 specially.  It has no way to tell what an
adress actually is.


That was the classic issue, cisco dropped in some code that dropped
routed traffic to .0 and .255 addrs.