The following text is copyright 2003 by Network World, permission is hearby given for reproduction, as long as attribution is given and this notice is included.

Re singing an old song?

By Scott Bradner

Just what is Sprint up to? With great fanfare Sprint announced on May 27th that "Sprint once again makes history with technology milestone." But this is an old, and confusing, story and the new announcement did nothing to make it a new story or to reduce the confusion. So why did Sprint make the announcement and why did the press fawn all over it?

The new Sprint press release (http://144.226.116.29/PR/CDA/PR_CDA_Press_Releases_Detail/1,3245,1111632,00.html) is almost the same as a press release from November 2001 (http://144.226.116.29/PR/CDA/PR_CDA_Press_Releases_Detail/0,3245,4081,00.html). Both releases talk about Sprint deciding to migrate their whole network over to a "packet network." But all may not be as it seems. It looks like the "packet network" Sprint is talking about is not IP, instead it is ATM, the telco dream network technology. The 2001 release is clearer about this than the 2003 one. The 2001 release says that the result of the aim is "an entire network evolution including Class 4 and Class 5 components to packet utilizing subscriber line over ATM (SloA) technology." The 2003 release is not quite so clear, it says that the Sprint will convert by "initially leveraging the high reliability of Asynchronous Transfer Mode (ATM)."

It looks to me that the releases, if not designed to be misleading, are at the very least very carefully worded. Most non-telco people in the data networking business would not immediately think of ATM when some company says that they are converting to a packet network. In ATM the min chunks of data are referred to as "cells" not "packets." One can make the argument that it is not technically incorrect to refer to cells as packets. I agree that is true, but I've rarely seen a case in the last few years when an ATM network has been referred to as a packet network unless the speaker is trying to make the reader incorrectly think of IP-based networks.

The new release is carefully worded, it says that the conversion "solidly positions Sprint for Internet Protocol (IP) applications." Someone skimming this announcement might assume that Sprint was saying that they would be able to run IP but that is not quite what it says.

A few years ago I said I would not do any more ATM-bashing columns. So, by assertion, this not an ATM-bashing column. But ATM is not IP. ATM is a circuit switch technology where the carrier determines what circuits are permitted to be setup. IP is a datagram protocol where the carrier forwards the packets towards the right destinations.

Why did Sprint issue the new release? Maybe because they are late in the original plan, or maybe because MCI has begun to say that they are going to actually converge their services over a real packet network, i.e., over IP and Sprint wanted to muddy the waters. Why did the press fawn so unquestioning over the announcement? I do not know, maybe you better ask them.

disclaimer: At Harvard, an unquestioning student is a failure, but the above questions are my own.