Tobias Oetiker <oetiker@dmu.ac.uk> said:
Tobias> The only problem I had so far, was that when I tried to
Tobias> discover our campus network, some hosts returned a netmask of
Tobias> 0.0.0.0 (eg kepler.dmu.ac.uk) These are PCs with trumpet
Tobias> winsock. They have their netmask configured correctly. But for
Tobias> some reason, they don't tell.
Tobias> Anyway, this prevented IP discover from producing a sencible
Tobias> map.
Tobias> My solution to the problem was, that I canged the discover
Tobias> module to assume a netmap of 255.255.255.0 when ever it finds
Tobias> a node returning 0.0.0.0
The idea of quessing a class C like network might work in many setups,
but not in all. There are some other options: Get the broken ICMP
implementation fixed. Well, this might not happen soon (BTW, cisco
router also do not respond to ICMP mask requests) but complaining
about it might help.
Another solution would be to add some SNMP probing code to query
ipAdEntNetMask. However, this is only of litte value if you do not
know the community strings (or the V2 configuration).
The current algorithm requires that at least one machine reports a
subnet mask. If there are machines which report different masks, the
majority will win (where all machines are ignored during the vote that
fail to report their mask). It would be possible to assume that
machines that do not report their mask will have a small mask if a
vote is started, thus giving some preference to small masks. But I
don't know if this will be a big win. Any comments?
Juergen