The following text is copyright 1993 by Network World, permission is hearby given for reproduction, as long as attribution is given and this notice is included.

If you ask "what application", you've already lost

By: Scott Bradner

A few years ago Harvard University participated in a trial of a Metropolitan Area Network (MAN) with the local regional telephone company. This was quite a nice package offering Ethernet speed LAN connections to four Boston & Cambridge sites. Before the trial started we had a few somewhat slower (2400 to 9600 baud) connections between most of the sites but the increase to 10 MB/sec made quite a difference. Before and during the trial the vendor sent up survey teams to ascertain what applications we were running over the test network. We kept telling them that while we were using dozens of network based applications (with a potential for hundreds) there were few, if any, additional applications enabled because of the higher speed pathways. Things simply took less time. The survey teams just could not understand and kept pumping us to tell them the two or three applications that could be used to justify a MAN service offering.

I'm telling this story in light of congressional testimony a few weeks ago pushing for the telecommunications vendors to provide the service that would be the foundation of the national data network infrastructure. I'm not sure that the time for that is here yet. The telecommunications industry is beginning to get the picture of what is involved in providing a inter-organizational data network. Unfortunately there are still some problems with their current understandings of the scope of the problem.

For example, the most common wide area data service offering from the telecommunications vendors is Virtual Private Networks (VPN). VPNs are a way for many branches of an organization to be connected but they are not pooled like the telephone system where one phone gets you to any other phone. Buying network connections with VPNs is like purchasing telephone service that allows you to talk to your branch office in NYC but not to some other business next door. You don't build an infrastructure by keeping the pile of parts separate. So far most of the offerings do not include any form of interconnection to what is known as the Internet. Sprint's SprintLink and Willtell's WillPower are two exceptions.

It is the mid-level networks, AlterNet, BARnet, NEARnet, SUREnet, PSInet, etc. that are actually providing the inter -organizational connectivity now and make up the bulk of the U.S. portion of the Internet. Most of these mid-levels are not intended just for woolly academicians and paid for by the feds. NEARnet (yes, I have a bias since I was one of the founders) started without NSF funds and is totally self supporting. Its membership is now made up of more than 66% commercial organizations. Whatever true infrastructure that does exist in this country is mostly due to these mid-levels which connect thousands of organizations together.

It may seem a bit silly that these networks purchase their communications links from a local telephone company, stick modems on the end and then resell the links to their customers. It might seem that one could eliminate the middle-man (the mid-level networks) and that the telephone companies could sell the conectivity service directly. Well, most of the problems that the mid-level networks are confronted with are not those of broken phone lines but those things that fall under the umbrella of "user services" and problem resolution. Helping the user configure their own environment to make use of, and protect themselves from, the Internet is not an easy task. The current telecommunications vendors have not yet understood this is a required part of an infrastructure service offering. (The current push for ISDN will not change the picture, it just eliminates the modems; the user service requirement is still there.)

It is reasonable to expect that the time will come when true interconnectivity and useful user support will be actually offered by the telecommunications vendors. Until then, it will be important that the creation of any national data infrastructure be done using the existing mid-level networks and under strong guidance or even with some service provision contracted for by the Federal Government. This will remain the case until the telecommunications vendors stop looking for the "killer application" and realize that it has been here all along -- it is called "connectivity".