On linux/unix, the configuration is the same.
What operating systems are your users on and how are they accessing hadoop? (Web Browser, command line, third party tool?)
Are your users logging into a domain (if they are on Windows?). If so, is that domain's realm the same one as your hadoop realm?
In order for us to answer your questions more precisely, we'll need to understand your intended use case.
Basically, though, there is no difference for how a client is configured and what requirements there are. Kerberos is a protocol, so, regardless of the type of server, clients' access to the KDC should be relatively the same.