Issue No. 03 - May/June (1999 vol. 3)
Every ten years or so, the most basic of telecommunications services, the telephone, has undergone a dramatic change. In the 1950s, the introduction of transatlantic coax cables allowed direct-dial international calls; in the 1960s, digital transmission and switching drastically improved the audio quality; in the 1970s, programmable switches enabled touch-tone dialing and local services such as call waiting; and in the 1980s, the widespread implementation of out-of-band common-channel signaling systems like Signaling System 7 (SS7) made services such as 800-numbers possible. These changes define a trajectory from analog transmission and signaling to digital, circuit-switched transmission and packet-based signaling.
In the 1990s, Internet telephony marks the latest step along this slow path to an all-packet infrastructure. Its history actually began 20 years earlier. The first papers on how to transmit voice were published in the early 1970s, and the first Internet packet audio experiments took place in August 1974, when real-time packet voice was demonstrated between the University of Southern California's Information Sciences Institute and the Massachusetts Institute of Technology's Lincoln Laboratory. The first Request for Comments for packet voice, RFC 741, was published in 1977.
Internet telephony developed relatively slowly until 1991 and 1992, when packet-audio experiments were performed on DARTnet, and the first IETF meetings were multicast across the Mbone. I made my first Internet phone call to Switzerland around 1992, but such communications were limited to research labs and targeted primarily at multiparty teleconferences, rather than person-to-person phone calls. Even if you had Internet connectivity of sufficient bandwidth, there was no way to reach a particular person short of exchanging each other's IP addresses by e-mail or regular telephone.
In 1995, Vocaltec introduced one of the first PC-based Internet telephony applications. Gateways to the public switched telephone system (PSTN) followed shortly thereafter, although most were initially limited to a few analog ports. There was also a shift away from using end systems like PCs, which connect directly to the Internet, and toward using regular telephones on both ends instead.
Convergence of Telecommunications and the Internet
This special issue offers a snapshot of some of the technologies for Internet telephony. Appropriate to the convergence of traditional "telecom" services and the Internet, the issue was solicited jointly by the IEEE Communications Society and the IEEE Computer Society, and six of the seven accepted articles appear in both IEEE Network and IEEE Internet Computing, magazines of the respective societies. This copublishing experiment reflects, I believe, the convergence between technologies and outlooks.
For the next few years, most Internet telephony users will employ either a gateway from a regular analog phone into the Internet or their home modem. An article by William Goodman, "Internet Telephony and Modem Delay," analyzes the factors that cause modems to add far more delay than simply the packet size divided by the line rate. The article illustrates how optimizing a component for high data throughput may inadvertently decrease its performance for a new service. Because it focuses on network optimizations, this article appears only in IEEE Network.
The article, "IETF Internet Telephony Architecture and Protocols," by Jonathan Rosenberg and myself, summarizes some of the protocols that have been developed or are emerging in this space, both for carrying the audio data and for establishing calls. IEEE Internet Computing is posting this article on its webzine, IC Online, as part of a special resource on Internet telephony (see "New Directions: Resource Pages @ IC Online," this issue, and http://computer.org/internet/telephony/).
Beyond carrying the human voice between two places, the telephone network has also generated a set of expectations. For example, we expect that the number shown on a caller-ID display is indeed correct or that the caller's anonymity is protected; we expect the quality of the call to be predictable and the 911 services to reach the fire and police departments instead of generating a message for a general protection fault: "Sorry, you have not configured your application properly." In the article, "Integration of Call Signaling and Resource Management for IP Telephony," Pawan Goyal et al. present a complete system architecture for Internet telephony that encompasses signaling for call setup, a protocol for unicast resource reservation, and a mechanism for moving capacity between the data and voice traffic sharing the same infrastructure.
One distinguishing factor of Internet telephony is that part of its behavior can be programmed by end users, not just by owners or switch manufacturers. Two articles investigate this issue from different angles. "A Voice over IP Service Architecture for Integrated Communications," by Daniele Rizzetto and Claudio Catania, illustrates how mobile Java code may be used to implement call routing and thus combine caller and callee preferences in a single node. "Programming Internet Telephony Services," by Jonathan Rosenberg, Jonathan Lennox, and myself, takes a somewhat different approach by proposing to extend the model used by Web services to access external resources, and applying it to one of the Internet telephony protocols. Just as an increasing number of people create their own Web pages or set up e-mail filters without being programmers, it is plausible that a large number of people will want to change the behavior of their phone in basic ways, for example, to increase their privacy from telemarketers during dinner. The article proposes a call processing language that is simple enough to be robust and predictable.
In environments like "residential gateways," which allow home users to plug their existing phones into a cable or asynchronous digital subscriber line (ADSL) modem, the service provider may want to maintain control over the services available to a subscriber and also over end-system costs. In the article, "An Architecture for Residential Internet Telephony Service," Christian Huitema et al. propose a protocol that allows a central controller to manage a large number of residential gateways. The same protocol can be used to manage large-scale gateways between circuit-switched networks and packet-switched networks, and possibly also between circuit-switched networks and "Internet phones," that is, appliances that have an Ethernet or ADSL connection. There are, of course, a host of operational issues that must be considered in moving toward ubiquitous Internet telephony.
The final article, "From POTS to PANS," by Christos A. Polyzois et al., identifies some of these issues, ranging from support of emergency calls to billing. It also raises the question of whether services should be implemented in the end system or in the devices owned by carriers.
Many aspects of 1990s Internet telephony might eventually seem as quaint as an 1890s crank phone. This decade may nevertheless be setting the foundation for version 2 of the most successful communications service in history.
Changing and upgrading computer equipment is hard, as the persistence of the 8086 architecture has shown. This difficulty pales, however, next to upgrading the worldwide network of 640 million telephones. We expect to upgrade our computer system at least every five years to accommodate new software, but we expect just as much that the rotary-dial phone installed by the Bell System in 1960 will connect to the most modern digital phone.
This is the predicament faced by Internet telephony. Indeed, a fundamental tension within the technical community might well be expressed as a matter of emphasis: one camp emphasizes the word "Internet"; the other, "telephony" (a distinction also crudely known as netheads versus bellheads). The latter group would prefer Internet telephony to be basically the well-known existing phone systems over packets, making a transition similar to the largely invisible upgrade from analog and coax or microwave to digital and fiber. The first group sees it largely as yet another application that runs on top of IP packets, with certain additional requirements such as quality of service and signaling.
Will Internet telephony and current Internet services share the Internet as "ships that pass in the night," with telephony just reusing the common packet transport infrastructure, or will it be the synchronous part of a multimedia communications architecture that encompasses Web and multimedia retrieval, e-mail and messaging, conferencing and collaboration, distributed games, and other applications yet to be invented? The challenge will be to realize the latter, while maintaining the ubiquity, universal connectivity, and reliability of the existing phone service.
On behalf of the IEEE Communications and Computer Societies, I thank the authors of the 25 papers submitted and the 67 reviewers who made this special issue possible. Particular thanks to the editorial board of IEEE Internet Computing for suggesting this exploration of converging technologies and to John Daigle and Jörg Liebeherr of IEEE Network who provided valuable short-notice feedback and editorial insight.
Henning Schulzrinne is an associate professor in the computer science and electrical engineering departments at Columbia University. His research interests encompass real-time multimedia network services in the Internet and modeling and performance evaluation. Schulzrinne received his undergraduate degree in economics and electrical engineering from the Technische Hochschule, Darmstadt, Germany, in 1984; his MSEE as a Fulbright scholar from the University of Cincinnati, Ohio, in 1987; and his PhD from the University of Massachusetts, Amherst, in 1992. He is on the editorial boards of the J. of Communications and Networks and IEEE Internet Computing. He co-chairs the IEEE Communications Society's Internet Technical Committee and is vice chair of its Technical Committee on Computer Communications. He is co-author of the Real-Time Protocol (RTP) for real-time Internet services, the signaling protocol for Internet multimedia conferences and telephony (SIP), and the stream control protocol for Internet media-on-demand (RTSP).