Server Push
Server push is a Netscape-only scheme for providing dynamic web content. The idea is that the web server sends content to the client with a special MIME type - multipart/x-mixed-replace - and specifies a "boundary string" that the browser looks for between the various parts of the output stream. It then displays the parts as they come over.
Some servers require special handling to ensure that the output stream isn't buffered on the server side- in the case of NCSA-derived servers that special handling is "non-parsed-headers." See the documentation for your server to determine if anything special is necessary for your own case.
The resources here, such as they are, were developed in the hopes of dreaming up a way to allow the content in the web browsers of remote students to mirror that of content in an instructor's web browser. Poor-man's distance learning.
A server-push example. (Netscape only.) Simply sends over a two-part document with a brief delay between parts.
The idea of this is that one user- an instructor?- browses around in the toy document tree (2), while other users- students?- run the script (1). That script monitors access times on a file that all the static documunts include (via server-side includes), and, when the script determines that the include file has just been read, it checks the access times on all the documents in the tree to guess which document the "instructor" has loaded. The script then sends that content to the "student." It uses server-push to do it. It's not really as good an idea as it sounds, in part because it scales up very poorly, and in part because it only works with Netscape, and in part because it isn't robust.