[Nottingham] Weird Stuff

Paul Sladen nottingham at mailman.lug.org.uk
Sat Sep 7 18:10:00 2002


On Fri, 6 Sep 2002, Robert Davies wrote:
>
> What's going to happen when someone injects packets destination
> 255.255.255.255 from source 255.255.255.255 into NTL's network on a
> variety of ports likely to have hosts responding?

``Not very much.''

> Surely they have no business routing this traffic.

You're quite right;  NTL (or anyone else) never *route* `255.255.255.255';
it is the limited-broadcast address and will never travel outside of a
single layer-2 domain (ie, it will be repeated (switched) but never routed).

DHCP relays proxy the BOOTP packets by altering them (they replace the
`0.0.0.0' and `255.255.255.255' addresses with the remote DHCP servers's and
their own Unicast addresses;  then do the inverse on the way back.
 
Wake-on-Lan also uses `255.255.255.255' so you may be able to have some fun
there...  (well it's actually just simple hardware that's interested in the
MAC address).

Like wifi in `infrastructure mode', chatter between nodes goes via the
central end-head to be re-broadcast out again so I suspect that the UBR may
not be relaying limited-broadcast packets as these would never normally
anything but `client -> headend' or `headend -> client'.

> It's fun enough on dialup when you get a rare '.255' address doled out to 
> you, you get to see how many chatty 'doze PCs are out there.

That's quite fine, if they're using prefixes shorter than /24 then .255
isn't always going to be a broadcast address;  or if they're using Pt-to-Pt
links and routeing /32s through the NAS platform then the issue of subnets
disappears completely.

...Presumabley 'doze boxes with broken PPP stacks.  :)

	-Paul
-- 
Nottingham, GB