Full marks to the other posts, but the advice about using "Content-Disposition" headers neglects Netscape 4.x browsers (at least in my experience). To get around this difference in browsers, I have used the following snippet of code, with credit going to Merlyn's web techniques column for the path-info-in-query-string trick:

# determine the http header to send the browser client:
my $user_agent = $q->user_agent; # CGI query object
if ($user_agent =~ m/MSIE/i) {        # for IE users, the following wo
+rks:
    print "Content-Disposition: attachment; filename = $name\n\n";
}
else {                                # for Netscape clients, need to 
+do a 2-pass trick here...
    if ($q->path_info() )    {
        # 2nd pass: collect the extra path info which Netscape client 
        # will use to name the downloaded file:
        print $q->header('application/x-octet-stream');
    }
    else {                        
        # 1st pass: add the name info to the http header path info var
        # and redirect back upon itself for the 2nd pass:
        $q->path_info("/$name");
        print $q->redirect($q->self_url());
        exit;
    }
}
[download]

Hope this helps

..Guv