Our succeessful experiences with Distributed Feature Composition have shown the value of building IP-based multimedia services compositionally.
SIP is now the dominant signaling protocol for building multimedia services on top of the Internet Protocol (IP). Because little is known about building SIP services compositionally, a team of AT&T researchers is working to adapt what we have learned from DFC, and make it available to the SIP community. The team consist of:
We are now converting our DFC implementation to a version that runs inside vendor-supplied SIP Servlet containers. As a first step, our ECharts language is a programming language based loosely on UML Statecharts, and the open-source release includes a SIP Servlet development kit.
We are also members of the Java Community Process JSR 289 expert group. This group is writing the next version of the SIP Servlet standard, which will include the DFC routing algorithm as the default method for composing SIP servlets.
When there is more than one application server in the signaling path between IP media endpoints, and the servers manipulate media flow, media flow must be controlled compositionally. To solve this problem, "Compositional control of IP media" presents an architecture-independent descriptive model, a set of high-level programming primitives, a formal specification of their compositional semantics, a signaling protocol, an implementation, and partial verification of correctness (Pamela Zave and Eric Cheung; IEEE Transactions on Software Engineering, to appear, 2008). The paper also discusses how the principles developed to solve this problem may help in making other network applications compositional. Here is a talk on compositional media control.
Recently we have designed an implemention of compositional media control, as specified above, in SIP. This work is described in "Generalized third-party call control in SIP networks" (Eric Cheung and Pamela Zave; Proceedings of the Second International Conference on Principles, Systems and Applications of IP Telecommunications, Springer-Verlag LNCS to appear, 2008). Although we have not yet deployed our implementation, analytic results suggest that the SIP implementation is significantly slower than our original design in "Compositional control of IP media". This latter paper explains the performance difference in terms of fundamental properties of the protocols used.
In telecommunications, audio signaling is the use of the audio channel for signaling and user-interface purposes. When features use audio signaling, and are assembled in a pipes-and-filters configuration, there is a potential for undesirable feature interactions. "Audio feature interactions in voice-over-IP" analyzes the potential feature interactions. It proposes a method for eliminating some of them, as well as directions for future work on the remaining interactions (Pamela Zave; Proceedings of the First International Conference on Principles, Systems and Applications of IP Telecommunications, 2007). The method can be implemented in SIP, using the techniques of compositional media control. Here is the conference talk on audio feature interactions.
"Understanding SIP through model-checking" reports on a project to build Promela models of SIP, and to verify them with the Spin model checker (Pamela Zave; Proceedings of the Second International Conference on Principles, Systems and Applications of IP Telecommunications, Springer-Verlag LNCS to appear, 2008). The paper discusses the following six models: