SIP SIMPLE Client SDK

<abbr title="SipDeveloperGuide*, depth=3">TOC</abbr>

= SIP SIMPLE SDK Integration Guide =

SIP SIMPLE client SDK is a Software Development Kit with a Python API designed for development of real-time communications end-points based on SIP and related protocols for multimedia like Audio, Instant Messaging, File Transfers, Desktop Sharing and Presence. Other media types can be added by using an extensible high-level API. The library has cross platform capabilities on Mac OSX, Microsoft Windows and Linux OS.

Motivation

SIP protocol is peer-to-peer by nature, each peer can be either client or server depending on direction of how session flows. The protocol can be built in end-points without the need for dedicated servers. This makes SIP the ideal candidate for adding real-time communications to self-organizing networks where the application runs in transitory end-points. This is not true for other protocols (e.g. XMPP) as they have a clear distinction of client and server roles which makes it unsuitable for self-organizing peer-to-peer networks. The second and most important feature is the SIP is agnostic to the type of media established between end-points. If two end-points decide to establish a certain media type over the network, no network element must understand that except for the concerned end-points themselves. For example audio is a most common media, known as VoIP, but adding another media type like chat, file transfer and desktop sharing is rather easy. As media planes, RTP (real-time protocol) and MSRP (message session relay protocol) have been defined for real time applications (e.g. audio and video) and reliable almost real-time streams (e.g IM and File Transfer) respectively. For both there are reliable NAT traversal mechanisms built-in.

State of the Art

The SDK is highly innovative. While many have done in various ways bits and pieces of the functions exposed by SIP SIMPLE client SDK, this toolkit is the only one refined for developing peers without the need of servers and contains all primitives that allows further development for additional of new functionality in a backwards compatible way. For example, the SDK provides all primitives to offer what Skype does in an integrated way, combining audio, chat, presence and file transfer in one logical session.

Current Status

SIP SIMPLE client SDK is reliable for production use and is used today in finished products that are downloaded hundreds of time per day. The products work on Linux, Mac OS, Windows desktop OS and Linux Servers.

Installation Instructions

SIP SIMPLE client SDK is available as debian package for the latest Ubuntu Linux distributions (Lucid, Maverick and Natty), Debian Linux distributions (Stable and Unstable) and can be manually installed on MacOSX 10.5, 10.6 and Microsoft Windows XP, Vista and 7 by following the documentation provided with the source code.

To install the SDK on Debian or Ubuntu, configure your deb repository as described [http://www.ag-projects.com/projects-products-96/683-software-repositories here], then:

{{{
sudo apt-get update
sudo apt-get install python-sipsimple
}}}

Detailed building instructions from source for the SDK and its dependencies are available [http://sipsimpleclient.com/wiki/SipInstallation here]

API Documentation * Online documentation is available at http://sipsimpleclient.com/wiki/SipMiddlewareApi * Printable documentation in PDF format is available at http://download.ag-projects.com/SipClient/SIPSIMPLE-Manual.pdf Usage Instructions

The SDK works on any platform that supports Python and provides direct access to the input and audio devices using one of the supported backends.

Using SIP SIMPLE client SDK library is no different than using any other Python library, after installing it in the system, one must import its modules into the program and use it. There is a [http://sipsimpleclient.com/wiki/SipMiddlewareApi#SIPApplication high level API] that automates and integrates most of the tedious tasks like starting and stopping all required sub-systems in the right order and performing most common applications like setting up an audio or a chat session so a developer can use the SDK by writing just a few lines of code without having to understand all the details or the used protocols.

To setup a simple audio call, print the connection state and hangup one has to write just a few lines of code, see Hello World example below.

A complete multi-party audio+chat+file transfer conferencing application has been written in 300 lines of code. On the other side of the scale developing a complete end-point like Skype is the most complex example of what the SDK may be used for. The complexity grows as one needs to interact with the end-user for more actions like creating conferences, interacting with telephony like applications, adding chat or handle presence updates. The complexity is dictated mainly by the external UI design, the library provides easy to understand high-level functions and the develeper does not need to understand SIP or other related protocols used inside the kit to achieve his goal.

=== Deployment Scenarios ===

The SDK can be used to build real-time communications end-points that operates in the following scenarios:

• On a LAN using Bonjour discovery mechanism for next-hop address resolution
• On the public Internet with any SIP Service provider using DNS for next-hop address resolution
• Part of a P2P overlay network like a DHT that provides routing and lookup services. For this, the SIP end-point built with the SDK will publish into the P2P overlay its Address of Records (AoR) address:port:protocol where it listens for incoming requests and will implement a lookup function to obtain the peer end-point addresses from the overlay when it wants to establish an outbound session.
• NAT traversal is done in the end-points using ICE methodology

All topologies can be combined together in the same client application.

Sample Code

=== Hello World Program ===

This is the hello world example that establishes a wideband audio session with a test number that plays a music tune. After installing the SDK, paste this code into a console and run it using Python.

{{{
#/usr/bin/python

from application.notification import NotificationCenter
from sipsimple.account import AccountManager
from sipsimple.application import SIPApplication
from sipsimple.core import SIPURI, ToHeader
from sipsimple.lookup import DNSLookup, DNSLookupError
from sipsimple.storage import FileStorage
from sipsimple.session import Session
from sipsimple.streams import AudioStream
from sipsimple.threading.green import run_in_green_thread
from threading import Event

class SimpleCallApplication(SIPApplication):

def init(self):
        SIPApplication.__init__(self)
        self.ended = Event()
        self.callee = None
        self.session = None
        notification_center = NotificationCenter()
        notification_center.add_observer(self)

def call(self, callee):
        self.callee = callee
        self.start(FileStorage('config'))

@run_in_green_thread
    def _NH_SIPApplicationDidStart(self, notification):
        self.callee = ToHeader(SIPURI.parse(self.callee))
        try:
            routes = DNSLookup().lookup_sip_proxy(self.callee.uri, ['udp']).wait()
        except DNSLookupError, e:
            print 'DNS lookup failed: %s' % str(e)
        else:
            account = AccountManager().default_account
            self.session = Session(account)
            self.session.connect(self.callee, routes, [AudioStream(account)])

def _NH_SIPSessionGotRingIndication(self, notification):
        print 'Ringing!'

def _NH_SIPSessionDidStart(self, notification):
        audio_stream = notification.data.streams[0]
        print 'Audio session established using "%s" codec at %sHz' % (audio_stream.codec, audio_stream.sample_rate)

def _NH_SIPSessionDidFail(self, notification):
        print 'Failed to connect'
        self.stop()

def _NH_SIPSessionDidEnd(self, notification):
        print 'Session ended'
        self.stop()

def _NH_SIPApplicationDidEnd(self, notification):
        self.ended.set()

1. place an audio call to the specified SIP URI in user@domain format
   target_uri="sip:3333@sip2sip.info"
   application = SimpleCallApplication()
   application.call(target_uri)
   print "Placing call to %s, press Enter to quit the program" % target_uri
   raw_input()
   application.session.end()
   application.ended.wait()
   }}}

=== Full Demo Programs ===

Fully featured sample programs are available in 'sipclients' package available in the [http://www.ag-projects.com/projects-products-96/683-software-repositories same repository] as python-sipsimple.

These programs operate in a Linux or MacOSX terminal and implement most of the functions provided by the the SDK.

{{{
sudo apt-get install sipclients
}}}

• sip-register - REGISTER a SIP end-point or detect Bonjour neighbors
• sip_audio_session - Setup a single SIP audio session using RTP/sRTP media
• sip_session - Complete client with multiple sessions for Audio, IM and File Transfer
• sip_message - Send and receive short messages using SIP MESSAGE method
• sip_publish_presence- PUBLISH presence information for a given SIP address
• sip_subscribe_winfo - SUBSCRIBE to the watcher list for given SIP address
• sip_subscribe_presence - SUBSCRIBE to Presence Event for a given SIP address
• sip_subscribe_rls - SUBSCRIBE for Presence Event to a list of SIP addresses
• sip_subscribe_mwi - SUBSCRIBE for Voicemail Message Waiting Indicator

For a description of the sample programs see http://sipsimpleclient.com/wiki/SipTesting

=== Finished Products ===

The SDK is used in several products since end of 2009. These are end-products that use the SDK and provide exhaustive examples for how the SDK was used to achieve the goals.

1. Multimedia SIP client for MacOS X available in the Mac App Store: http://itunes.apple.com/us/app/blink-pro/id404360415?mt=12&ls=1 
 1. VoIP Client for Linux: Blink for Ubuntu, and Debian, use same repositories from AG Projects and do 'sudo apt-get install blink' 
 1. VoIP client for Windows: https://blink.sipthor.net/download.phtml?download&os=nt
 1. Multiparty Conference Application with Wideband Audio, IM chat and File Transfers: http://sylkserver.com

Non-Python environments

As the programming choice was Python, the SDK will appeal to people that want to develop applications in Python.

If one want to use the library into another environment it must check if is feasible or not as embedding a high-level programming language into another is not a straight forward process if at all possible.

=== Cocoa Objective C ===

MacOS platform is using Objective C for native development. But it also provide a bridge between Python and Objective C. This makes it possible to write pure Python programs that run into the native OS without much changes.

http://pyobjc.sourceforge.net/

The SDK was fully used without any problems to build Blink for MacOSX by using this bridge.

=== Qt Framework ===

Qt Framework from Nokia (formerly Trolltech) is using C for native development. But it also has Python bindings. This makes it possible to write pure Python programs that integrate with Qt framework.

http://qt.nokia.com/

The SDK was fully used without any problems to build Blink for Windows and Linux by using Qt Framework and its Python bindings.

=== Web Browser Plugin ===

Today, web browsers do not give direct access to the input and output audio devices to their plugins, which is required by a real-time audio application. Browsers do not support direct embedding of Python code either. If one looks for example at how Google Talk plugin in the browser works, it actually installs a regular server daemon into the system that performs similar functions to SIP SIMPLE Client SDK and the web browser interacts with it by using a proprietary API running on the same host between the daemon and the browser plugin. The software that handles the media, signaling, integration with the audio device does not run in the browser itself and the browser is used only as a GUI.

There is a recent initiative to form a work group in both W3C and IETF to address this important issue of accessing and processing audio data within the browser itself as today this is not possible. This requires cooperation from the browser manufacturers, third-party developers cannot do this on their own without cooperation from the web browser makers (like Adobe has obtained for its Flash product). This initiative is called RTCweb:

http://rtc-web.alvestrand.com

Google published a first specification and code drop in May 2011 and IETF and W3C had contacts to setup the future agenda.

SIP SIMPLE Client SDK cannot run in the browser today for the reasons highlighted above as browsers won't support Python nor direct access to audio devices both required by the SDK to operate. However one could build a program that stays behind the browser in the same way as Google Talk daemon does.

=== C Language ===

How to possibly integrate Python into C is described here:

http://docs.python.org/extending/embedding.html

=== Java Language ===

There is a bridge between Python and Java.

http://www.jython.org/

=== Translating into other languages ===

Python is a suitable language for designing complex state machines. The state machine for a multimedia real-time application (SIP or other protocols) is orders of magnitude more complex than a near-real-time file transfer protocols like BitTorent. Real time communications imply using both signaling and media that may travel over different paths and their combined state must be aggregated in the end-points.

• SIP signaling has 10 states with 16 possible transitions between them, a diagram is [http://sipsimpleclient.com/raw-attachment/wiki/SipCoreApiDocumentation/sipsimple-core-invite-state-machine.png here]
• RTP media used for audio and video has 6 to 10 states depending on the type of media
• MSRP media used for chat functionality has multiple states
• ICE NAT traversal has many states interrelated with both SIP signaling and RTP media plane
• Presence Publish/Subscribe/Notify mechanism has multiple states

One must describe the transitions of states, which may contain complex data and can be raised in different threads. There are multiple sockets in use using several transports. When all combined together the final application is quite complex and achieving the SDK functionality in a low level programming language like C would be a major undertaking.