You should take a look at IO::Socket, or if your familiar with sockets in C (and don't want to do it the easy way), then take a look at Socket.pm
Also, if you considering handling it without forking, take a look at IO::Select (again, if you're a glutton for punishment, just use select and have fun with your vectors)
As to whether multiplexing, or forking is better, it kinda depends on preference, and on the application. I generally prefer to do forking when I have my choice or either will do.
When doing forking, you need to decide if you want to fork as needed, or do preforking (similar to what Apache normally does), also you'll want to keep track of your children (you don't want to fill up your process table do you?). Also, don't forget to set $SIG{CHLD} properly to deal with dead children

I personally feel that forking on demand is the easiest set up, but multiplexing can give you some tighter control (for instance, if for each connection, you'll be accessing a single resource constantly, a multiplexed server would be easiest to eliminate contention)

I believe there are also some good examples in the Perl Cookbook, and in Advanced Perl Programming (both from O'Reilly of course)

Hope that helps,
cephas