Mangement vlan - restrict to each building and don't span across all sites. Main reason is STP. If you accidentally create a loop in youtr mangement work then any broadcast storms would be limited to that building rather than spanning the entire campus setup.
Private addressing. To be honest it's more important to summarise effectively than worry about using every bit of a subnet with VLSM. As you say it's not as though you are going to run out. Allocating a block of subnets to each building and then summarising out is worth doing though.
dsitribution switches - depends on how you want to do it. Usually they are interconnected either with L3 or more commonly L2 link. Assuming your distribution switches are doing the inter-vlan routing for the access-layer then a L2 link is the most common for HSRP. Note though if all access-layer switches are uplinked to both switches then you can use a L3 interconnect between your distribution switches and the HSRP messages simply go via the access-layer switches.
Like i say L2 is more common still, but L3 can be useful for the interconnect if your distro switches support GLBP. Reason being that with L2 interconnect one of the uplinks has to block per vlan. With L3 interconnect both L2 uplinks from the access-layer switch can be forwarding because there is no STP loop and both distro switches can share the load from clients.
Where they are not connected i would guess the routing for the vlans would either -
1) be per distro switch which would mean each distro switch is independant. Never really seen this though and the switch would need to be running redundant sups etc. in my opinion
2) routing for vlans takes place at core. This would also mean distro to core would need to be L2. Again haven't really seen this nowadays and not recommended.
3) routing for vlans in access-layer which means distro switches really just act as routers. Not really seen that setup either but i can see how it would make sense ie. again each access-layer switch would have dual uplinks and they could both be forwarding traffic. If a distro switch failed then the other link is there anyway. Like i say, they really become routers to all intents and purposes.
This also supposes that any other devices beyond the distro switches such as routers are connected to both distro switches. If the router was only connected to one distro switch how would the other distro switch ever get the routes ? Suppose they could peer via the access-layer switches but you really don't want to do that, very bad design.
So in summary haven't really seen many examples (in real networks) of non-interconnected distribution switches
Jon
