Wednesday, November 12, 2003

Telirati Newsletter #24

Here is an idea whose time didn't come, but is ripe for a second look soon. The success of Linux in embedded and consumer electronics applications means that many of these appliances will grow to become the personal servers described here. The general stability of file sharing protocols has made them possible to reverse engineer, and Macs, Windows PCs, and Linux nodes will like happily together.

Telirati Newsletter #24: The next battle: Personal networks, tiny servers

A lot of the smoke and noise coming from the computer battlefield over the next three years will be over servers and networks. For most, the perception will be that this is all about servers for the un-exploited small and medium sized business market, and about the further downsizing of formerly mainframe-based systems. These are, certainly, important developments. But like a really big rock rolling downhill, there are not going to be many surprises about where all this is going.

There is, however, a market in a wonderful state of ferment: personal networks and really small servers. Big servers are capable of a wide range of purposes: file sharing, backup, faxing, telephony, etc., even if the each server specializes on one such task. Small servers will have a fixed purpose. They are built for a single task, or related set of tasks. Their hardware is designed for those tasks and no others.

Who will buy these small servers? What person needs a network? Good questions and ones that do not necessarily have answers that are optimal for the interested parties. So we start with those asking the questions: PC makers who have seen the price of a typical PC erode. Previously, PCs got vastly more powerful without dropping from the $2000 to $3000 price range. But the next level of PC penetration and the popularity of the Internet ended this comfortable situation. PCs have fallen to, and through, the $1000 level. Which has led to a hope, a scheme, a theory, under which the size, power, and total price of an individual PC customer’s purchase can be lifted to higher, more profit-yielding levels. This theory: the personal network.

This is a term you probably have not heard. Personal networks sound a bit self-contradictory. They are not fully personal, but they cannot require the level of management networks require now.

Some of the problems that have to be solved before personal networks will work well:

User management. I don’t manage how my wife and daughter use their computers. Their Macintoshes and my PCs do not inter-operate to any extent except to share a LaserWriter, which happily operates on an AppleTalk network and on a parallel port connection to a PC, no further configuration required than to plug both connections in. We don’t share an Internet connection, and the effort of acquiring a high-speed connection and sharing it would be too high, although it is possible to do. Personal networks must accommodate the style and practice of personal computer use in order to become manageable and applicable to commonplace problems. Currently, PC users accommodate networks in order to use them. Personal computers are just getting over the inherent complexity of being completely general-purpose machines. While one can expect the first personal networks to be made in the image of existing data networks, this will have to change, otherwise personal networks will remain geekware.

Simplification of communications. Shared Internet access is the driving force behind personal networks. It is all people really care about. Everything else is in a distant second place. Personal networks and small servers must provide a solution to this problem in order to be viable. At first, the mainstream solution will rely on multiple modem connections. Then personal networks will have to transition to enabling sharing of high speed connections, as these finally become widely available.

Current market segmentation. Even if networks can be made simple, there are forces operating in the market to prevent this happening. The tide sweeping over the industry is one in which small businesses are being networked. This is a hugely profitable trend. It is driving the near and mid-term profit expectations of Microsoft, and of the server operations of PC makers. Altering the price and value composition of servers and networks, and tampering with the channel for delivering these products is considered dangerous. The computer industry will have to learn a new type of market segmentation, and accept some cannibalization of its small business market in order to open a new mass market among individuals and the smallest businesses.

If these difficulties (and other outside the scope of this missive) are overcome, we will see a world where all devices, such as printers, become print servers, and where a set-top box or “home router� also becomes a communications server and the convergence point of home computing. The personal network may have one server, say something like a multifunction printer/fax/scanner with a brain, or there may be several specialized servers. These servers will have to learn to cooperate and stay out of each others’ way, to figure out which function (DHCP, for example) is the responsibility of which server, and enable management that does become a geometrically growing burden as the number of small servers increases. In this world, the concept of a home computer will be replaced by multiple computers that truly are personal, a concept worth keeping in the front of one’s mind: PCs are more like underwear and toothbrushes than they are like the silverware.

So if this is the future, where is telephony in this picture? This is what must be considered: if the conventional small business key system looks out of place next to today’s PCs, how dislocated must it seem next to a future of pervasive networks. A fundamental task for the telephony industry will be to come up with telephony CPE that fits, exploits, extends, and simplifies this coming reality.

Copyright 1998 Zigurd Mednieks. May be reproduced and redistributed with attribution and this notice intact.