The basic difference between those two approaches is where the
"browsing smarts" reside -- client or server. The path of your
explorations suggests some of the pros and cons:
Pros for the server-side approach: the browsing interface can be
customized for the topic at hand and, more importantly, it can be
instantly upgraded without all the distribution problems of ordinary
software. (Presumably these reasons attracted you to the server
approach in the first place.)
Cons: server-side smarts can be a heavy burden to both server and
client. The server must maintain "account" information for large
numbers of users, even though the majority will make only brief use
of the system. And the user has the burden of trying to keep track
of all the accounts on different servers.
In other words, both pure-client and pure-server approaches have
serious disadvantages. What is really needed is a client/server
balance.
For example, if the amount of "user customization/preference" info is
not overly large (up to a few 10K), server-side smarts can still be
made efficient by relying on the client to store the user-specific
info between sessions. This can be done conveniently with current
Web tools by providing users with a log-on page (laced with hidden
form fields) that they can save on their own machines, and then
resubmit to start a new session. This allows the server to only
maintain short-lived "session" state, rather than long-lived "user"
state, and so is much more efficient.
Of course, more sophisticated client/server balancing is possible.
--------------------------------------------------------------------
Paul Burchard <burchard@math.utah.edu>
``I'm still learning how to count backwards from infinity...''
--------------------------------------------------------------------