Sunday, January 18, 2004

Telirati Newsletter #34

Four years ago I warned that computers would have to simultaneously continue to acquire the capabilities of large-scale computing systems, and they would simultaneously have to get easier to operate, maintain, and defend against cracking. There has been precious little progress since then: PCs can't be easily clustered, despite the potential .NET has to enable pervasive distributed processing. Security from attacks is still elusive. Encypted file systems are still an obscure feature. An Internet of servers is as far away as ever. And the prospects are still dim. Microsoft has distracted itself with DRM while Linux prepares to make they jump to desktops.

Telirati Newsletter #34: A long way to go

According to a recent report in The Register, a researcher at Sony devised a novel way of designating files to be copied from one computer to another: He used a pointing device, a pen in this case, to indicate which file he wanted to transfer on the monitor attached to the computer on which the file resides, and then using the same pen, to indicate which folder the file should be copied to on a destination computer. This very simple improvement illustrates just how far we have yet to go:

How will multiple computing devices be made workable for individual users, or informal, unmanaged groups of computer users such as families sharing two or three (or more, as in my case) computers at home?

What is the essence of “personal” in personal computers? Wholly possessing your computer has enormous benefits. So does merging its environment with that of other computers on a network. While it is a simple, almost trivial, idea to enable one pointing device to operate on more than one computer, consider the implementation details – how is this to be made easy and intuitive? (Light-pens have certain properties that may make it relatively easy to implement such shared use, but that does not make pens the best choice for users.) Currently, notions of network, computer, state, environment, etc. are far too crude to form a virtual environment as fluid and intuitive as our physical environment.

Sharing a computer is needlessly complex because current concepts of security and resource sharing developed in an era where a person made it his profession to manage the arrangements for sharing. This is no longer economically sound in an era of $1000 servers.

An individual with multiple computers is in even worse shape. Some efforts to manage portable environments has been made for networks of computers, but, as with servers, if computers cost $500, why not have three of them? Why not, but for the fact it is a pain in the behind to keep three computers in order. Document synchronization software is a venerable genre of utility, but a built-in system-wide solution is the only really satisfactory way forward. I should be able to say that one PC is a slave to, or subset of, another, and specify the character and behavior of subset in a simple fashion.

These laments are not mere carping. I get paid to mess around with computers. For most people, the messing around is a dire cost. What does this mean? It means we are not nearly done with the evolution of personal computing environments.

Another measure of the extent to which we are not done is the extent to which concepts from large-scale computing have been absorbed by personal computers. A networking software company recently tried to bring the concept of clustering to PCs. Not servers, but desktop PCs. Why not? If clustering is good for servers, if it makes servers more reliable and more “available” it should do the same for a network of workstations, as they are dynamically added or taken away from a network, data should not become inaccessible, and aggregate computing power should expand as more computers are networked. Unfortunately they only got as far as clustering storage, and had to try to sell the idea as an interesting type of peer-to-peer NOS.

PCs will have to simultaneously continue absorbing the stuff of “serious” computing, and keep getting simpler to operate. Increasingly, there isn’t any computing more “serious” than what can be accomplished with a cluster of PCs. Increasingly, the only administrator is you. Security, privacy protection, redundancy, automated vigilance against cracking are all going to have to become simple, possibly invisible, parts of personal computing in order to enable PCs to unlock the power of pervasive networking.

With wireless networking, you may find yourself in the presence of networks all the time. Some might be your neighbors’ networks! If only to prevent, on a large and horrifying scale, the equivalent embarrassment of overhearing one’s neighbors on the nursery intercom, managing what is personal and what is connected will have to advance with more than incremental steps. Or, to put it in context, it would be embarrassing if “Furby” toys were more adept at self-configuration when networked than a PC is.

Copyright 1998 Zigurd Mednieks.