In the race to understand IP telephony, it is critical that net professionals' knowledge remain ahead of vendor hype. The last thing we need is vendors telling us how to define and evaluate new technology. For most, that means opening up the black box known as the PBX, understanding traffic characteristics of packet voice and taking a first look at unifying directory, voice and e-mail systems.

I find it strange that the people who brought us plain old telephone service (POTS) had to call corporate phone systems a Private Branch Exchange. These devices are changing so much today that I propose we give them a new name. I call them modular, open phone systems, or MOPS, for short.

Vendors are deconstructing the black-box PBXs and reimplementing the features almost helter-skelter across varying combinations of open and closed hardware and software platforms. Network managers immediately need to understand the critical elements that comprise MOPS and begin understanding the finer points of each.

For example, you'll find three vendors that refer to their systems as NT-based. While that statement would lead one to believe that the systems are directly comparable, that is hardly the case. One might have all the phone system hardware and call-processing logic built into a stand-alone (proprietary) box and simply use NT for system configuration and user administration. The next might have the phone hardware implemented as PCI boards that slot into the NT server and all call-processing functions implemented as NT services. A third might have all its phone hardware and processing logic built into a PCI board that simply resides in an NT Server and draws power from the bus - not relying at all on NT for any services.

Each of these can technically be called NT-based, but each will have radically different dependencies on NT. More importantly, the fault tolerance (or lack thereof) of NT has a dramatically different effect on each.

With the addition of real-time voice traffic, designers will need to take a much closer look at the impact of device and network latency on audio quality as well as the impact of data traffic spikes (congestion) on the conversation. While this will be more of an issue on the WAN, it cannot be ignored on the LAN.

Finally, net managers will have to start looking at what I call "back-end convergence" - the linking of directories and deliverables of voice and data systems.

Called unified messaging by many vendors, it covers advances such as letting your phone system access your NT user directory instead of maintaining its own, duplicate directory of users. Equally interesting is the move by some vendors to deliver your voice mail as e-mail-attached .WAV files. Nice thought, but who wants to download a dozen or more huge audio files when checking e-mail from a 56K-bps hotel dial-up connection?

Where this back-end convergence makes sense, net professionals have to make sure that they are not locking themselves into proprietary links.

Tolly is president of The Tolly Group, a strategic consulting and independent testing firm in Manasquan, N.J. He can be reached at ktolly@tolly.com or www.tolly.com.