TCP/IP - very weird default gateway problem



Hi there,

we are having a very weird "problem" with a TCP/IP configuration on our W2K
Pro machines. I stumbled on it while giving a course on classless IP
addressing. Here's the description of the setup I'm using and the weirdest
phenomena I've seen so far.

I have a small LAN of 11 PCs, 5 of them running W2K Pro and the others
running XP Pro. A router is present on the LAN to connect our little network
to other classrooms.

We use the following IP settings on one of the PCs running W2K : IP address, subnetmask, Default Gateway; So
far so good, no problem, both PC and Default Gateway reside on the same
subnet. To simulate a problem we change the subnetmask to
This causes the PC and the Default gateway to find themselves in different
subnets. The PC should have problems and should not be able to gain access
to another LAN since it can not find it's own default gateway.

If you take a look at the binary picture : 224 = 11100000, thus 3 bits are
used for additional subnetting.

Now if I compare the first 3 bits of the PC and the default gateway (201 =
11001001, 250 = 11111010) then the logical conclusion would be that the PC
can not reach his own default gateway. At least that is the way an XP
machine reacts when we fill in this configuration. And that is how it should

For some unknown, obscure reason, all the W2K machines still manage to find
their default gateway and even more, can gain access to the other

I've tried about everything I could think of, rebooted the machines, cleared
the ARP cache tables, used Ethereal to capture and disect every packet the
PC sends... but to no avail.

What is even worse, I can recreate the problem on an XP machine. If I first
disable the network card, then enter the same configuration (host IP that
does not reside on the same subnet as its default gateway) and then
re-enable the network card, the XP machine will find its default gateway,
which logically is impossible. However, if I correct the configuration (host
IP and default gateway back in the same subnet) and afterwards try to
re-enter the "faulty" adresses, the XP machines react exactly as they
should. At that moment they don't find their default gateways which is a
perfectly normal situation.

My question to the world : does anyone know what is going on here ? Is there
some "automatic function" in W2K that transparently "corrects" a faulty IP
configuration ? If so, is it possible to disable this function and how do I
do that ?

please post an answer in this newsgroup, maybe it can help others, or mail
me at (e-mail address removed)

thank you, my eternal gratitude awaits you if you can give me an answer !



This is a "feature" of Windows. Although the linux types hate it (it defies
the rules), it is extremely useful. I discovered this (quite by accident as
you did) in my MCSE class and actually used it to my advantage when setting
up a routed network (3 subnets) where the requirements were that the subnets
couldn't communicate with each other, but they all needed Internet access
via the same (SOHO) router. Using a combination of VLANs, three network
cards in the DC, and some DNS tweaks to get the domain SRV records to
resolve to the correct subnet, we achieved (for a really low cost) three
subnets that couldn't see each other, couldn't route to each other - but all
had broadcast access to the DC and Internet via the same router. IMHO, MS
did this one right.



To get your problem to work (fail?), you'll need an IP address outside of
the router's mask.

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question