pfsockopen() works great on IIS/Windows7 installation, it keeps connection open which is good for performance. However, there is one caveat: when connection is broken because of physical net failure, pfsockopen() returns handle as if connection was working. Subsequent call to fwrite() returns false so you have information about error. The problem is that after physical net connection is restored the situation doesn't change: pfsockopen() still returns handle and fwrite() returns false. In other words, PHP sticks to old connection that is not working (if you use fsockopen() instead, it will connect properly). Situation goes back to normal after 30 minutes when PHP closes unused connection.
The solution to this problem is to call fclose() on socket handle when fwrite() returns false.

To see if it's really a new connection, or a reused one, you can use ftell() - and see if ther's been any traffic on the connection. If it's more than 0, then it's a reused connection.

Persistent connections either in socket or databases should be used only in servers where the limits are well defined. For example, the number of allowed connections in a database must be greater than the number of Apache's processes, or the connections will be refused by the database (this will surely occur if you use persistent connections). The same may occur with socket connections. This is up to the service configuration. In my opinion, persistent connections are useful only if you have total control over the one or more servers involved, like on a heavy loaded dedicated server for example, where the little gain in performance worth the use of such connections. Never use them in a shared server.

OK, WRT to the p* functions opening a new connection when one already exists. It is my understanting that (under Apache anyways) this is on a per-process basis. If you do a 'ps auxw|grep httpd' on your server you will see more than one process. What p* does is make a p-connection on one of those processes only, the one that actually handles your request. Chances are that when you hit the page again it will be answered by a different process. I'm guessing if you keep hitting reload you'll get around to the original process again and there will be no error message or second connection open. Anyhow, this is true of all p* functions; they open not one connection per server, but one connection per server _process_.

Here is how to POST a form action to a SSL server's cgi and retrieve output with pfsockopen

<?php

$host = gethostbyaddr($_SERVER['REMOTE_ADDR']);

# working vars
$host = 'www.example.com';
$service_uri = '/cgi-bin/processACT';
$vars ='code=22&act=TEST';

# compose HTTP request header
$header = "Host: $host\r\n";
$header .= "User-Agent: PHP Script\r\n";
$header .= "Content-Type: application/x-www-form-urlencoded\r\n";
$header .= "Content-Length: ".strlen($vars)."\r\n";
$header .= "Connection: close\r\n\r\n";

$fp = pfsockopen("ssl://".$host, 443, $errno, $errstr);
if (!$fp) {
   echo "$errstr ($errno)<br/>\n";
   echo $fp;
} else {
    fputs($fp, "POST $service_uri  HTTP/1.1\r\n");
    fputs($fp, $header.$vars);
    fwrite($fp, $out);
    while (!feof($fp)) {
        echo fgets($fp, 128);
    }
    fclose($fp);
}

?>