A question came up on the comp.lang.tcl newsgroup about a CGI which gave continuous updates to a Netscape browser. The original question had to do with a problem where Netscape stayed busy even after the CGI was finished (which was about two hours(!) after its started). This evolved into a discussion of how to get page updates that would mimic a real-time monitor without resorting to the use of Java.
JavaScript provides a way to modify the contents of a DHTML page, and it occured to me that by sending small JavaScript segments to the browser, it might be possible to have it appear like a real-time monitor that could add or update sections of page rather than simply append to a page.
I've written a couple of proof-of-concept pages that do this via PHP[L1 ], but any CGI-like backend will work. What counts is the contents of the JavaScript and the browser's DOM. You can see the working examples and the PHP code which produced them at http://www.rlenter.com/stuff/server-monitor.html . The critical part seems to be
Here is a complete example which replaces the second paragraph (identified via the id=x1 tag) with new information which is then updated periodically. You can view the page via http://www.rlenter.com/stuff/server-mon2.html . e
<?php // -*- php -*-
$header = '
<!DOCTYPE html PUBLIC "-//W3C//DTD xhtml 1.0 Transitional//EN"
"http://www.w3.org/TR/REC-html40/loose.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en">
<head>
<title>JavaScript Monitor Test</title>
</head>
<body>
<h1>JavaScript Monitor Test</h1>
<p> This is a test of a JavaScript Monitor which feeds additional
information to the page in "real-time" and eventually finishes the
page. This particular "monitor" writes out the page then the
replaces a paragraph with updated information using JavaScript. If
you look at the page source in your browser, you will see the
original HTML source plus all of the JavaScript commands. The
"trick" here is that the JavaScript parts which update the page come
out slowly. <a href="server-mon2.phps">Click here</a> to see the
PHP source for this page.</p>
<p id ="x1">Replace me.</p>
<hr />
<address>
Roland B. Roberts, PhD<br />
RL Enterprises
</address>
</body>
';
$trailer = '</html>';
echo eval ('?> ' . $header . '<?php ');
echo '<script language="JavaScript" type="text/javascript">' . "\n";
echo ' myP=document.getElementById("x1");' . "\n";
echo ' myTextNode=myP.childNodes[0];' . "\n";
echo '</script>' . "\n";
flush ();
for ($i = 1; $i <= 10; $i++) {
sleep (2);
echo '<script>' . "\n";
if ($i == 1) {
echo ' myTextNode.data = "' . "$i" . '";' . "\n";
} else {
echo ' myTextNode.data = myTextNode.data + "' . "..$i" . '";' . "\n";
}
echo '</script>' . "\n";
flush ();
}
echo $trailer;
?>
Still to do: identify the versions of DOM, JavaScript, and CSS that make this possible, and correlate them with delivered browser versions. Also, CL is preparing demonstrations written in (server-side) Tcl [L2 ], and working on others which update images rather than text.
[CL needs to explain some of the applications of the technique.]
Examples currently work with: IE 5.5 and 6, Netscape 6, Mozilla 0.9.5, ... ...
Examples do not work with: Netscape 4, Opera 6, ...
Arjen Markus I have been experimenting with something akin to this - using the browser as a device for "on-line visualisation". That is, a long-running computation produces an output file, a plotting program produces a GIF-file from the last record in the output file and the browser regularly checks if there is a new picture. It is not fool-proof as yet, but the principle seems to work.
CL responds that, yes, there are both browser-pull and server-push applications of the technique. Its kernel, to my mind, has to do with DOM navigation or computation. A few more comments appear in [L3 ].