I may be trying to teach Granny to suck eggs here but you cannot flout the
rules.
Subnet masks must always be a series of 1's then 0's. You cannot mix and
match. The mask is used to separate the host and network portions ie the
last non-zero octet must always be one of 192, 224, 240, 248 or 252 (254 is
allowed in Class A in the second or third octet or Class B in the third
octet but not in either case the fourth octet). 128 is also not allowed to
be the next octet from the standard netmask for each class in subnetting.
All the following netmasks allow no hosts
Class A netmasks 255.255.255.254 or 255.128.0.0.0
Class B netmasks 255.255.255.254 or 255.255.128.0
Class C netmasks 255.255.255.254 or 255.255.255.128
The netmask 255.255.255.255 is used to signify this host on this network.
Class A networks 1.0.0.0 to 126.0.0.0, may have netmasks from 255.0.0.0 to
255.255.255.252
(Network 127.0.0.0 being special. Note it has 2 to the power 24 minus 2
hosts! What a waste.)
Class B networks 128.0.0.0 to 191.255.0.0, may have netmasks 255.255.0.0 to
255.255.255.252
Class C networks 192.0.0.0 to 239.255.255.0, may have netmasks 255.255.255.0
to 255.255.255.252
(in the three Classes above, note the exclusion of the 128 subnet mask)
You cannot use a range of netmasks on a LAN. You can use multiple netmasks
if each netmask is used on its own LAN separated by a router or routers or
switches using VLANs.
On top of all this you cannot use the first and last host in any network or
subnet as these are the wire and broadcast addresses.
Private networks Class A 10.x.x.x, Class B 172.31.x.x or Class C 192.168.x.x
NT does not care because it is probably using NetBIOS and WINS and not
TCP/IP and DNS
Pinging the network card in the host will always work, if the drivers are
set up correctly and the card works, because the host knows its MAC address
and does not require to do an ARP request. In your situation I believe the
ARP requests are malfunctioning due to a bad netmask. Therefore no ping to
hosts on the LAN with the malformed netmask.
When bits are given back in a netmask rather than being borrowed to create
subnets, it is called supernetting. Routers have to be specially configured
to allow this to work.
Cisco routers can use the bottom network in any range of subnets but they
have to have this feature turned on (ie you can also use 128 in the next
octet in the standard netmask for each class or 254 in the fourth octet but
it is a waste of IP address ranges). Legacy routers will throw a wobbly if
you use the first subnet in any range of subnets, routing would not work.
This can be a bit of a minefield but our Cisco students do get it in the
end.
Regards
Bill