The components shown in Figure 1 interoperate along the control plane through a collection of interfaces. Of particular interest to the media service application developer are the interfaces between the application server and the media server. These APIs can be divided into three categories: signaling, connection control, and media control.

As was previously mentioned, signaling messages are used to establish and manage the media session between the end-user terminal and the media server. The primary signaling interfaces for next-generation applications are SIP for IP and IMS and Signaling System 7 (SS7). Intel NetStructure^® SS7 Boards provide applications with basic and advanced signaling capabilities in both wireline and wireless networks. For SIP signaling, applications can use either SIP protocol stacks from third parties or Intel's own APIs. Intel NetStructure Host Media Processing Software supports standard SIP protocols through the Global Call API. This is discussed in more detail below.

Connection control APIs allow the application to control the routing and flow of media along the data plane. There are numerous methods for controlling the media data flow. For the IP-centric next-generation media server, the most common method involves directing Real-Time Protocol (RTP) streams using SIP. The Media Gateway Control Protocol (MGCP) [7], as defined by the Internet Engineering Task Force (IETF), allows applications to control the routing and flow of media through a gateway.

The category of interface most relevant to this discussion, media control, covers a wide variety of media functions. It includes basic voice messaging such as the ability to play announcements and record messages, as well as video functions. Media control also covers the ability to accept control input from the end user. This input may be in the form of in-band tones such as DTMF or speech, as well as out-of-band messages such as those defined by RFC-2833 [8]. Intel supports a variety of APIs for media control. In the following paragraphs we discuss a few notable ones.

Voice Extensible Markup Language (VoiceXML) [9] is a standard XML-based language that enables voice/speech interactions with Web pages. VoiceXML is defined by the World Wide Web Consortium (W3C). Application logic is expressed in the form of VoiceXML documents that are processed by a VoiceXML interpreter (a.k.a. browser) which, in turn, drives the media server. VoiceXML is popular for Interactive Voice Response (IVR) applications because it defines a simple yet powerful high-level programming language.

Two additional media control interfaces attracting interest for next-generation media servers are the Media Sessions Markup Language (MSML) [10] and Media Objects Markup Language (MOML) [11]. MSML and MOML are Internet drafts submitted to the SIPPING work group of the IETF. MSML/MOML are XML-based interfaces that, while capable of standing on their own, are designed to complement one another. In contrast to VoiceXML, which presents a high-level language, MSML/MOML defines a framework for describing a wide range of media functions in an extensible manner. Besides offering rich semantics, MSML/MOML is attractive for use by next-generation media servers because it is designed to be compatible with SIP. The downside of MSML and MOML is that they are not yet recognized as standards and are still relatively immature.

Another notable media control interface is Intel's R4 "C" language APIs. The R4 APIs have been a staple of computer telephony applications programming for over 10 years. They are supported on Intel media processing technology boards as well as Intel NetStructure Host Media Processing Software.

All of the interfaces described above can be used with Intel-based media servers. Some, such as VoiceXML, can be acquired through third parties, while others are directly available from Intel. In the following sections we discuss Intel NetStructure Host Media Processing Software and the R4 APIs. We demonstrate how to implement a simple media service application that will run on media servers built using HMP software.