One might then use one CGI script to do logging, another to do
authentication, and then the final one to do a search (for instance).
Different subtrees could have their own logging/authentication/etc
properties. Use kerberos here, but IP address there. Offer automatic
decompression on these files, but not on those.
Basically, the server implementation is getting spread over URL space.
Server simplicity, modularity, and service flexibility bought with
system load (multiple script processes).
Just a thought,