Strictly speaking, "pure" client-server model assumes that the server cannot send notification; it's a model where the server play purely passive role: everything is initialized by one or another client, and the server only responds to requests. Elapsed time is totally irrelevant here. This model is extremely weak, limiting and inefficient for many applications; I never appreciated it, but it it became almost dominating.
For some background, please read:
http://en.wikipedia.org/wiki/Pull_technology[
^],
http://en.wikipedia.org/wiki/Push_technology[
^],
http://en.wikipedia.org/wiki/Client%E2%80%93server_model[
^].
At the same time, there are way more advanced and useful models, such as
publisher-subscriber and other.
How about the Web? Unfortunately, it is primarily pull technology (and for some basic functionality, like the one needed for decent Web chat applications, this is a big obstacle), but the doors are not completely closed, for those who can invest in more complexity. Please read on "server push" in the second article referenced above. These days, one popular framework called SignalR can help to work around the problem:
http://en.wikipedia.org/wiki/SignalR[
^],
http://www.asp.net/signalr[
^].
See also my past answers on related topics:
Application 'dashboard' for website accounts[
^],
an amateur question in socket programming[
^],
Multple clients from same port Number[
^].
—SA