如何使用PHP套接字获取HTTP响应

Question

I need to write a PHP script which "executes" HTTP requests using a local proxy (squid). The HTTP request is received from the client, the script sends the request to the proxy, receives the HTTP response from the proxy and returns it to the client after some processing. I know all the bad things you're going to tell me about this approach but trust me, I need to do it.

I'm using PHP sockets to accomplish that. This is a fragment of the code where I try to relay the request (it's based on an example I found on the web):

<?php 
$PROXY="localhost";
$PORT="3128";
$BUFFER_SIZE=1024*1024*5; //5Mb buffer

// (An untested HTTP request example, replace if it is not correct pls)
$request = "GET http://www.google.com HTTP/1.1\nHost: www.google.com\n\n";

// Create socket
if(!($sock = socket_create(AF_INET, SOCK_STREAM, 0)))
{
    $errorcode = socket_last_error();
    $errormsg = socket_strerror($errorcode);
    die("Couldn't create socket: [$errorcode] $errormsg\n");
}

//Connect socket to proxy server
if(!socket_connect($sock , $HOST , $PORT))
{
    $errorcode = socket_last_error();
    $errormsg = socket_strerror($errorcode);
    die("Could not connect: [$errorcode] $errormsg\n");
}

//Send the message to the proxy
if( ! socket_send ( $sock , $request, strlen($request) , 0))
{
    $errorcode = socket_last_error();
    $errormsg = socket_strerror($errorcode); 
    die("Could not send data: [$errorcode] $errormsg\n");
}

//Now receive reply from proxy
if(socket_recv ( $sock , $buf , $BUFFER_SIZE , MSG_WAITALL ) === FALSE)
{
    $errorcode = socket_last_error();
    $errormsg = socket_strerror($errorcode); 
    die("Could not receive data: [$errorcode] $errormsg\n");
}

//print the received message
print $buf;

This code sends the message to the proxy, but... it gets blocked. After some research I've found what is the problem: the script waits until at least BUFFER_SIZE bytes are received or the connection is closed.

I can avoid this setting a timeout option:

if(!socket_set_option($sock,SOL_SOCKET,SO_RCVTIMEO,array("sec"=>2, "usec"=>0)))
{   $errorcode = socket_last_error();
    $errormsg = socket_strerror($errorcode);
    die("Can't set options: [$errorcode] $errormsg\n");
};

But this doesn't solve my problem, because I want to receive the whole HTTP response, and it the request can be HTTP 1.1

I've been thinking about it, and the only idea I have is to write code to analize the HTTP response to see where it ends. I've been searching some PHP code which does that but I haven't found anything.

So the questions I have are:

• Is there a way to store the whole HTTP response into a buffer without decoding it?
• Is there an easy way to decode an HTTP response with PHP?
• Is there an easier way to send and receive request to a proxy without using sockets?

Thanks a lot.

Update:

I've tried a loop, as recommended by Maskime. This is the code:

while($response=socket_recv ( $sock , $buf , 1 , MSG_WAITALL ))
{
    ...
    print $buf;
} 

It reads data and exits when the timeout arrives, not before. I can't use this approach because if I set a timeout all the request will suffer at least that delay, and if I set it too low some requests wont have time to load.

Answer 1

After some investigation, it seemed that you need to "decode" the HTTP/1.1 response to handle it correctly. I haven't found a library to do that as I needed, so I tried another approach.

What I've done is to modify the requests at proxy level so the connections became non-persistent. To do that, you need to set this directives in squid.conf (perhaps only a subset of them is enough, but I haven't tried):

client_persistent_connections off
server_persistent_connections off
persistent_connection_after_error off

With that configuration, it works as I wanted. It's a good idea to adjust the receive timeout to more than two seconds, or a lot of requests will be lost. Ten seconds is working fine for me, but it could be more: it will raise only when there is a real timeout.

I would like to know the response to my second question, if you have an idea it will be good to know it. I've been looking at Snoopy library but it works with HTTP/1.0 and have not investigated any further.