or
It's Deja Vu All Over Again
by Adrian Berry
Twenty years ago there were two essentially separate industries involved in getting electronic information from point 'A' to point 'B'. These groups were divided not so much by the type of information but by the type of circuit used to deliver it. The traditional methods involved point-to-point circuits with a single information stream, including things like telephone, telegraph, telex and facsimile. On the other hand, moving information between computers involved multiple streams of information over a single circuit which may have been either point-to-point or multidrop (multiple connections on the same physical medium). The practitioners in both of these separate fields had very different experiences and methodologies, and for the most part kept their distance - we had 'data guys' and 'voice guys' (no gender bias intended).
In 1968 the US Federal Communications Commission passed the famous 'Carterfone' decision, which in effect allowed privately owned telecommunications devices to be attached to the Public Switched Telephone Network or PSTN (the fact that the intent of the Carterfone was to allow the use of Ham Radios to bypass long-distance charges is another story). By the mid 1980s in North America there were several manufacturers of telephone equipment selling systems to private businesses for connection to telephone lines from the telephone company. The most advanced of these systems had now progressed from manipulating analog signals to converting all traffic within the local system to a digital signal and processing the bit stream through what would now be considered a crude computer. The capacities of these systems (Private Branch Exchanges, or PBXs) were measured by the amount of
information throughput of which they were capable, usually a measurement called a CCS (centi-call second; 36 CCS is the equivalent of 3600 seconds of traffic per hour, or one hour per hour). But one company in California which manufactured computer systems for the US Department of Defence decided to build a better PBX, and of course being 'data guys' measured it's performance not in CCS but in MIPS. A certain large data company in Armonk, NY, was so impressed that they bought the company and started to push all of their customers (the 'data guys', of course, not the 'voice guys') to buy their new product. By the end of the decade the writing was on the wall for separate voice departments.
But regardless of reporting structure, the 'voice guys' and the 'data guys' remained in separate disciplines and did their own thing. Experiments in carrying voice traffic over packet networks were not tremendously successful; industry had a very large investment in circuit-switched technology that was supposed to last ten to fifteen years; and IBM eventually admitted defeat and closed down their voice operation. Meanwhile the 'data guys' had enough to worry about, what with this newfangled 'Web' thing and the now-soaring demand for electronic mail. So voice was once again left alone to do its own thing.
But as the capacity of the Internet grew, so did its capability to transmit any form of information. At the same time, advances in encoding schemes reduced the bandwidth required to carry 'toll-quality' telephone traffic (quality here is a relative term, by the way - digital telephony signal quality is 4kHz bandwidth and around 45dB S/N ratio, just like an analogue phone line, which is not exactly hi-fi). This, along with incredible advances in computing power, allowed the application of the principles developed from the ARPANet "Network Voice Protocol" experiments dating back to 1973, which had previously required massively-parallel processing machines to run. And so VoIP was 'born' - or perhaps delivered. It is still in its infancy, but the writing is now on the wall not only for the separate voice and data departments, but also for the remnants of the circuit-switched
industry. Currently the major manufacturers of these dinosaurs can produce fairly convincing interface components to attach IP-based voice products to their 'legacy' systems, but at some point the ROI just doesn't make sense. VoIP solutions shouldn't even require special hardware; a server is a server, whatever format the information it is processing. Granted, at some point there has to be a bridge device to the PSTN to talk to the outside world, but sometime soon the common carriers are going to convert to VoIP themselves - it just makes good business sense.
Of course this is going to mean quite an upheaval for the small companies who have built an industry out of providing devices which connect to Plain Old Telephone Service (POTS) lines. Whether it is an answering machine or a conferencing system, it will now need to be redesigned to allow for VoIP access. There is nothing quite like informing the Director of IT that s/he is going to have to provide you with an Analogue Terminal Adapter to connect the conferencing system for the Boardroom to his/her brand-new Cisco VoIP telephone system. This problem is something of a repeat of the introduction of ISDN, or Integrated Services Digital Network, by the telephone companies in the late 1980s. ISDN at that time was presented as the wave of the future; all telephone service was going to be digital in the developed world within 10 years. It never happened, but a lot of companies spent a
fortune trying to develop products that would connect to everybody's proprietary interpretation of the ISDN signalling protocol. If it didn't exist before, ISDN most definitely contributed to the rise of the concept of the 'superset' of a given standard.
But this time around it doesn't look quite so bad - although it's not perfect. There seems to be some consensus on the adoption of the Session Initiation Protocol, or SIP, for most VoIP terminal communication. SIP was developed by the Internet Engineering Task Force for use with all types of multimedia sessions such as online gaming. It is claimed to be far simpler to implement than the main alternative which is H.323 (developed by the International Telecommunications Union - "The Phone Company"). It looks at this point as if simpler endpoint devices (IP telephones) will utilise SIP signalling, while more complex devices such as Videoconference codecs will tend to use H.323 signalling (which they do now). And SIP has a tremendous Open Source following, so it is relatively straightforward for designers and engineers to find information on how to produce a SIP-compliant device.
And VoIP signal quality, although starting out at exactly the same point as previous technologies, does have allowance built in for higher bitrates and resolutions.
And of course the neat thing is that now anyone with an Internet connection and a multimedia computer can connect to anyone elsewhere in the world with the same thing and talk - and possibly see as well - without incurring per-minute charges. Arthur C. Clarke predicted the death of long-distance charges by the year 2000; he was a little optimistic, but we may yet see it in our lifetimes.
SIP Foundry Open Source Community
