draft-test-latest.md

---
title: STUN/TURN using PHP in Despair
abbrev: STuPiD
docname: draft-hartke-xmpp-stupid-latest
date: 2009-07-05
category: info

ipr: trust200902
area: General
workgroup: XMPP Working Group
keyword: Internet-Draft

stand_alone: yes
pi: [toc, sortrefs, symrefs]

author:
 -
    ins: K. Hartke
    name: Klaus Hartke
    organization: Universität Bremen TZI
    email: hartke@tzi.org
 -
    ins: C. Bormann
    name: Carsten Bormann
    org: Universität Bremen TZI
    street: Postfach 330440
    city: Bremen
    code: D-28359
    country: Germany
    phone: +49-421-218-63921
    facsimile: +49-421-218-7000
    email: cabo@tzi.org

normative:
  RFC2119:
  RFC3986:
  RFC4086:
  RFC4648:

informative:
  RFC5389:
  I-D.ietf-behave-turn:
  STUNT:
    target: http://deusty.blogspot.com/2007/09/stunt-out-of-band-channels.html
    title: STUNT & out-of-band channels
    author:
      name: Robbie Hanson
      ins: R. Hanson
    date: 2007-09-17
  I-D.meyer-xmpp-e2e-encryption:
  I-D.ietf-xmpp-3920bis:



--- abstract

NAT (Network Address Translator) Traversal may require TURN
(Traversal Using Relays around NAT) functionality in certain
cases that are not unlikely to occur.  There is little
incentive to deploy TURN servers, except by those who need
them — who may not be in a position to deploy a new protocol
on an Internet-connected node, in particular not one with
deployment requirements as high as those of TURN.

"STUN/TURN using PHP in Despair" is a highly deployable
protocol for obtaining TURN-like functionality, while also
providing the most important function of STUN.

--- middle

Introduction        {#problems}
============

NAT (Network Address Translator) Traversal may require TURN
(Traversal Using Relays around NAT)
{{I-D.ietf-behave-turn}}
functionality in certain
cases that are not unlikely to occur.  There is little
incentive to deploy TURN servers, except by those who need
them — who may not be in a position to deploy a new protocol
on an Internet-connected node, in particular not one with
deployment requirements as high as those of TURN.

"STUN/TURN using PHP in Despair" is a highly deployable
protocol for obtaining TURN-like functionality, while also
providing the most important function of STUN
{{RFC5389}}.

The high degree of deployability is achieved by making STuPiD
a Web service, implementable in any Web application deployment
scheme.  As PHP appears to be the solution of choice for
avoiding deployment problems in the Web world, a PHP-based
sample implementation of STuPiD is presented in {{figimpl}} in {{impl}}.
(This single-page script has been tested with a free-of-charge
web hoster, so it should be deployable by literally everyone.)


The Need for Standardization   {#need}
----------------------------

If STuPiD is so easy to deploy, why standardize on it?
First of all, STuPiD server implementations will be done by
other people than the clients making use of the service.
Clearly communicating between these communities is a good
idea, in particular if there are security considerations.

Having one standard form of STuPiD service instead of one
specific to each kind of client also creates an incentive
for optimized implementations.

Finally, where STuPiD becomes part of a client standard
(such as a potential extension to XMPP's in-band byte-stream
protocol as hinted in {{xmpp}}), it is a good
thing if STuPiD is already defined.

Hence, this document focuses on the definition of the STuPiD
service itself, tries to make this as general as possible
without increasing complexity or cost and leaves the details
of any client standards to future documents.


Basic Protocol Operation   {#ops}
========================

The STuPiD protocol will typically be used with application
instances that first attempt to obtain connectivity using
mechanisms similar to those described in the STUN
specification {{RFC5389}}.  However, with STuPiD,
STUN is not really needed for TCP, as was demonstrated in
previous STUN-like implementations {{STUNT}}.
Instead, STuPiD (like {{STUNT}}) provides a
simple Web service that
echoes the remote address and port of an incoming HTTP
request; in most cases, this is enough to get the job done.

In case no connection can be established with this simple
STUN(T)-like mechanism, a TURN-like relay is needed as a final
fall-back.
The STuPiD protocol supports this, but solely provides a way
for the data to be
relayed.  STuPiD relies on an out-of-band channel to notify
the peer whenever new data is available (synchronization signal).
See {{xmpp}} for one likely example of such an
out-of-band channel.
(Note that the out-of-band channel may have a much lower
throughput than the STuPiD relay channel — this is exactly
the case in the example provided in {{xmpp}},
where the out-of-band channel is typically throughput-limited
to on the order of a few kilobits per second.)

By designing the STuPiD web service in such a way that it can
be implemented by a simple PHP script such as that presented
in {{impl}}, it is easy to deploy by those who
need the STuPiD services.
The combination of the low-throughput out-of-band channel for
synchronization and the STuPiD web service for bulk data
relaying is somewhat silly but gets the job done.

The STuPiD data relay is implemented as follows (see {{figops}}):

1. Peer A, the source of the data to be relayed, stores a chunk of
   data at the STuPiD server using an opaque identifier, the "chunk
   identifier". How that chunk identifier is chosen is local to Peer
   A; it could be composed of a random session id and a sequence number.

2. Peer A notifies the receiver of the data, Peer
   B, that a new data chunk is available, specifying the URI needed
   for retrieval.
   This notification is provided through an out-out-band channel.
   (As an optimization for multiple consecutive transfers, A might
   inform B once of a constant prefix of that URI and only send a
   varying part such as a sequence number in each notification —
   this is something to be decided in the client-client notification
   protocol.)

3. Peer B retrieves the data from the STuPiD server using the URI
   provided by Peer A.

Note that the data transfer mechanism is one-way, i.e. to send
data in the other direction as well, Peer B needs to perform
the same steps using a STuPiD server at the same or a
different location.

~~~~~~~~~~


        STuPiD   ```````````````````````````````,
        Script   <----------------------------. ,
                                              | ,
          ^ ,                                 | ,
          | ,                                 | ,
    (1)   | ,                                 | ,  (3)
    POST  | ,                                 | ,  GET
          | ,                                 | ,
          | v                                 | v

        Peer A   ----------------------->   Peer B
                           (2)
                       out-of-band
                       Notification
~~~~~~~~~~
{: #figops title="STuPiD Protocol Operation"}


Protocol Definition
===================

Terminology          {#Terminology}
-----------
In this document, the key words "MUST", "MUST NOT", "REQUIRED",
"SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY",
and "OPTIONAL" are to be interpreted as described in BCP 14, RFC 2119
{{RFC2119}} and indicate requirement levels for compliant STuPiD
implementations.


Discovering External IP Address and Port
----------------------------------------

A client may discover its external IP address and the port
required for port prediction by performing a HTTP GET
request to a STuPiD server. The STuPiD server MUST reply
with the remote address and remote port in the following
format:

host ":" port

where 'host' and 'port' are defined as in {{RFC3986}}.


Storing Data
------------

Data chunks are stored using the POST request of HTTP. The
STuPiD server MUST support one URI parameter which is passed
as query-string:

'chid':  A unique ID identifying the data chunk to be stored.
The ID SHOULD be chosen from the characters of the base64url
set {{RFC4648}}.

The payload of the POST request MUST be the data to be
stored. The 'Content-Type' SHOULD be
'application/octet-stream', although a STuPiD server
implementation SHOULD simply ignore the 'Content-Type' as a
client implementation may be restricted and may not able to
specify a specific 'Content-Type'.  (E.g., in certain cases,
the peer may be limited to sending the data as
multipart-form-encoded — still, the data is stored as a
byte stream.)

STuPiD servers may reject data chunks that are larger than
some predefined limit.
This maximum size in bytes of each data chunk is RECOMMENDED
to be 65536 or more.

As HTTP already provides data transparency,
the data chunk SHOULD NOT be encoded using Base64 or any
other data transparency mechanism; in any case, the STuPiD
server will not attempt to decode the chunk.

The sender MUST wait for the HTTP response before
going on to notify the receiver.


Notification
------------

The sender notifies the receiver of the data chunk by passing
via an out-of-band channel (which is not part of the STuPiD
protocol):

The full URL from which the data chunk can be retrieved,
i.e. the same URL that was used to store the data chunk,
including the chunk ID parameter.

The exact notification mechanism over the out-of-band channel
and the definition of a session is dependent on the
out-of-band channel.  See {{xmpp}} for one
example of such an out-of-band channel.


Retrieving Data
---------------

The notified peer retrieves the data chunk using a GET request
with the URL supplied by the sender. The STuPiD server MUST
set the 'Content-Type' of the returned body to
'application/octet-stream'.


Implementation Notes
====================

A STuPiD server implementation SHOULD delete stored data some
time after it was stored. It is RECOMMENDED not to delete the
data before five minutes have elapsed after it was stored.
Different client protocols will have different reactions to
data that have been deleted prematurely and cannot be
retrieved by the notified peer; this may be as trivial as
packet loss or it may cause a reliable byte-stream to fail
({{impl}}).
(TODO: It may be useful to provide some hints in the storing
POST request.)

STuPiD clients should aggregate data in order to minimize the
number of requests to the STuPiD server per second.
The specific aggregation method chosen depends on the data
rate required (and the maximum chunk size), the latency
requirements, and the application semantics.

Clearly, it is up to the implementation to decide how the data
chunks are actually stored.  A sufficiently silly STuPiD server
implementation might for instance use a MySQL database.


Security Considerations
=======================

The security objectives of STuPiD are to be as secure as if
NAT traversal had succeeded, i.e., an on-path attacker can
overhear and fake messages, but an off-path attacker cannot.
If a higher level of security is desired, it should be
provided on top of the data relayed by STuPiD, e.g. by using
XTLS {{I-D.meyer-xmpp-e2e-encryption}}.

Much of the security of STuPiD is based on the assumption that
an off-path attacker cannot guess the chunk identifiers.  A
suitable source of randomness {{RFC4086}} should
be used to generate at least a sufficiently large part of the
chunk identifiers (e.g., the chunk identifier could be a hard
to guess prefix followed by a serial number).

To protect the STuPiD server against denial of service and
possibly some forms of theft of service, it is RECOMMENDED
that the POST side of the STuPiD server be protected by some
form of authentication such as HTTP authentication.  There is
little need to protect the GET side.

--- back


Examples  {#xmp}
========

This appendix provides some examples of the STuPiD protocol operation.

~~~~~~~~~~
   Request:

      GET /stupid.php HTTP/1.0
      User-Agent: Example/1.11.4
      Accept: */*
      Host: example.org
      Connection: Keep-Alive

   Response:

      HTTP/1.1 200 OK
      Date: Sun, 05 Jul 2009 00:30:37 GMT
      Server: Apache/2.2
      Cache-Control: no-cache, must-revalidate
      Expires: Sat, 26 Jul 1997 05:00:00 GMT
      Vary: Accept-Encoding
      Content-Length: 17
      Keep-Alive: timeout=1, max=400
      Connection: Keep-Alive
      Content-Type: application/octet-stream

      192.0.2.239:36654
~~~~~~~~~~
{: #figxmpdisco title="Discovering External IP Address and Port"}

~~~~~~~~~~
   Request:

      POST /stupid.php?chid=i781hf64-0 HTTP/1.0
      User-Agent: Example/1.11.4
      Accept: */*
      Host: example.org
      Connection: Keep-Alive
      Content-Type: application/octet-stream
      Content-Length: 11

      Hello World

   Response:

      HTTP/1.1 200 OK
      Date: Sun, 05 Jul 2009 00:20:34 GMT
      Server: Apache/2.2
      Cache-Control: no-cache, must-revalidate
      Expires: Sat, 26 Jul 1997 05:00:00 GMT
      Vary: Accept-Encoding
      Content-Length: 0
      Keep-Alive: timeout=1, max=400
      Connection: Keep-Alive
      Content-Type: application/octet-stream
~~~~~~~~~~
{: #figxmpstore title="Storing Data"}

~~~~~~~~~~
   Request:

      GET /stupid.php?chid=i781hf64-0 HTTP/1.0
      User-Agent: Example/1.11.4
      Accept: */*
      Host: example.org
      Connection: Keep-Alive

   Response:

      HTTP/1.1 200 OK
      Date: Sun, 05 Jul 2009 00:21:29 GMT
      Server: Apache/2.2
      Cache-Control: no-cache, must-revalidate
      Expires: Sat, 26 Jul 1997 05:00:00 GMT
      Vary: Accept-Encoding
      Content-Length: 11
      Keep-Alive: timeout=1, max=400
      Connection: Keep-Alive
      Content-Type: application/octet-stream

      Hello World
~~~~~~~~~~
{: #figxmpretr title="Retrieving Data"}


Sample Implementation     {#impl}
=====================

~~~~~~~~~~
<?php
header("Cache-Control: no-cache, must-revalidate");
header("Expires: Sat, 26 Jul 1997 05:00:00 GMT");
header("Content-Type: application/octet-stream");

mysql_connect(localhost, "username", "password");
mysql_select_db("stupid");

$chid = mysql_real_escape_string($_GET["chid"]);

if ($_SERVER["REQUEST_METHOD"] == "GET") {
   if (empty($chid)) {
      echo $_SERVER["REMOTE_ADDR"] . ":" . $_SERVER["REMOTE_PORT"];
   } elseif ($result = mysql_query("SELECT `data` FROM `Data` " .
                         "WHERE `chid` = '$chid'")) {
      if ($row = mysql_fetch_array($result, MYSQL_ASSOC)) {
         echo base64_decode($row["data"]);
      } else {
         header("HTTP/1.0 404 Not Found");
      }
      mysql_free_result($result);
   } else {
      header("HTTP/1.0 404 Not Found");
   }
} elseif ($_SERVER["REQUEST_METHOD"] == "POST") {
   if (empty($chid)) {
      header("HTTP/1.0 404 Not Found");
   } else {
      mysql_query("DELETE FROM `Data` " .
                  "WHERE `timestamp` < DATE_SUB(NOW(), INTERVAL 5 MINUTE)");
      $data = base64_encode(file_get_contents("php://input"));
      if (!mysql_query("INSERT INTO `Data` (`chid`, `data`) " .
                       "VALUES ('$chid', '$data')")) {
         header("HTTP/1.0 403 Bad Request");
      }
   }
} else {
   header("HTTP/1.0 405 Method Not Allowed");
   header("Allow: GET, HEAD, POST");
}
mysql_close();
?>
~~~~~~~~~~
{: #figimpl title="STuPiD Sample Implementation"}


Using XMPP as Out-Of-Band Channel  {#xmpp}
=================================

XMPP {{I-D.ietf-xmpp-3920bis}} is a good choice for
an out-of-band channel.

The notification protocol is closely modeled after XMPP's
In-Band Bytestreams (IBB, see
http://xmpp.org/extensions/xep-0047.html). Just replace the
namespace and insert the STuPiD Retrieval URI instead of the
actual Base64 encoded data, see {{figxmpnots}}.
(Note that the current proposal redundantly sends a sid and a
seq as well as the chid composed of these two; it may be
possible to optimize this, possibly sending the constant prefix
of the URI once at bytestream creation time.)

Notifications MUST be processed in the order they are
received. If an out-of-sequence notification is received for a
particular session (determined by checking the 'seq' attribute),
then this indicates that a notification has been lost. The
recipient MUST NOT process such an out-of-sequence notification,
nor any that follow it within the same session; instead, the
recipient MUST consider the session invalid.  (Adapted from
http://xmpp.org/extensions/xep-0047.html#send)

Of course, other methods can be used for setup and teardown, such as Jingle
(see http://xmpp.org/extensions/xep-0261.html).

~~~~~~~~~~
      <iq from='romeo@montague.net/orchard'
          id='jn3h8g65'
          to='juliet@capulet.com/balcony'
          type='set'>
        <open xmlns='urn:xmpp:tmp:stupid'
              block-size='65536'
              sid='i781hf64'
              stanza='iq'/>
      </iq>
~~~~~~~~~~
{: #figxmpcri title="Creating a Bytestream: Initiator requests session"}


~~~~~~~~~~
      <iq from='juliet@capulet.com/balcony'
          id='jn3h8g65'
          to='romeo@montague.net/orchard'
          type='result'/>
~~~~~~~~~~
{: #figxmpcrr title="Creating a Bytestream: Responder accepts session"}



~~~~~~~~~~
      <iq from='romeo@montague.net/orchard'
          id='kr91n475'
          to='juliet@capulet.com/balcony'
          type='set'>
        <data xmlns='urn:xmpp:tmp:stupid'
              seq='0'
              sid='i781hf64'
              url='http://example.org/stupid.php?chid=i781hf64-0'/>
      </iq>
~~~~~~~~~~
{: #figxmpnots title="Sending Notifications: Notification in an IQ stanza"}

~~~~~~~~~~
      <iq from='juliet@capulet.com/balcony'
          id='kr91n475'
          to='romeo@montague.net/orchard'
          type='result'/>
~~~~~~~~~~
{: #figxmpnota title="Sending Notifications: Acknowledging notification using IQ"}

~~~~~~~~~~
      <iq from='romeo@montague.net/orchard'
          id='us71g45j'
          to='juliet@capulet.com/balcony'
          type='set'>
        <close xmlns='urn:xmpp:tmp:stupid'
               sid='i781hf64'/>
      </iq>
~~~~~~~~~~
{: #figxmpclor title="Closing the Bytestream: Request"}

~~~~~~~~~~
      <iq from='juliet@capulet.com/balcony'
          id='us71g45j'
          to='romeo@montague.net/orchard'
          type='result'/>
~~~~~~~~~~
{: #figxmpclos title="Closing the Bytestream: Success response"}