Question : Superscope on a multinet

If you want to have multiple subnets on a single subnet (ie 1 ethernet switch) I presume you create 2 scopes and add them into a superscope. My question however is how do the clients/server know which addresses should be assigned to which clients? Do you use user options or something? Otherwise it would be random as to whether a client ends up on a particular subnet wouldn't it?
Thanks for any input.

Answer : Superscope on a multinet

> multiple subnets on a single subnet (ie 1 ethernet switch) I presume you create 2 scopes and add them into a superscope. My question however is how do the clients/server know which addresses should be assigned to which clients?

Simple answer. Everyone is on the same wire. DHCP is broadcast, the client does not need to know topology. The server will hand out addresses first come first served. My observation was that it at least used to hand them out in sequence, from the beginning of the superscope to the end (even if some in the middle were released).

> Do you use user options or something?

No

> Otherwise it would be random as to whether a client ends up on a particular subnet wouldn't it?

Yes. From the user perspective, the resulting subnet is unpredicatble. Viewing from the server, however, as above, I have seen it as predictable.

>  But if the allocation is still random then why bother with a superscope.

Two problems. One is, that many products implementing servers will not permit a scope to be larger than ~256. Two, is that your ranges may not be in sequence. So the super one effectively can combine ranges that have little in common.

> I was thinking the idea was to have separate networks on a single ethernet segment.

no

>  For security for example

For this what you can do is to reserve ranges, or specify in policy what range is for what. Possibly a specifc scope, or substantial portion could be dedicated to non-workstations, and more than routers you can add servers and printers, and, even some workstations to a dedicated address that will never change, even though they will continue to request it.

>  But if the allocation is random then it seems I have the wrong idea about the purpose of superscopes.

Maybe. Randomness is the nature of dhcp, although the client can request some specific criteria be met. The purpose of superscope is really no more than permitting a larger number of addresses to be in the common pool than there would otherwise be. As this is NT TA, let me add, I also recommend the server to a ping prior to allocating addresses. Most disagree with me, thinking it too much traffic, but I claim the packets are really small and infrequent, and there is value to be had in ensuring that no one has the address already (ex: hard-coded, or roque, or simple mistake). This feature was unavailable in MS prior to development of superscopes, even though it is an independent issue.