The Inevitable, Vanishing Personal Computer.

The Vision

It is inevitable that the physical manifestations of personal computers will vanish into the surrounding 
infrastructure, into the very fabric of buildings, clothes, vehicles and so forth . As this happens, 
computer scientists will need to come to terms with systems that interact in an ad hoc manner with 
their users, as well as with each other. Concepts of input, output, persistence of information and 
computation will alter. The notion of ownership of information and computation will drastically alter, 
along a trajectory that is only partially visible today.

In this context, the challenge is across a broad range of computer science, including both systems and 
users. We will require a coherent approach to the networking, operating systems, programming 
environment for applications, interfaces and users. Such a system must accommodate completely novel 
structures such as unreliable components, and yet appear as reliable and useable as todays' PCs, or 
hopefully far more so!

The Trends

Between 1980 and 1995, a £1000 computer evolved from having "3Ms" (1 Megabyte of memory, 1 
MIPS processing, 1 Mbps of networking; even 4 Ms if you include 1 Mpixel of display) to "3Gs" 
(Gigabyte of memory, Ghz processor, Gbps networking, even 3D graphics, etc). This is now far more 
than enough for most sensible personal computing needs, especially now, given the ability to provide 
services within the network. 

An apparent trend in computer science is that we will soon produce constant performance at a falling 
price on the same curve, leading to the facilities in a 2002 desktop being available for under £1 by 
2012. There is evidence that this direction is already being followed to some extent, with PDAs, 
mobile cell phones and hand held gaming computers falling below the £100 mark. However a lot of this 
is achieved through reduced functionality. 

By analogy with the "4Ms", "3Gs", we can envisage the "5Cs" (cubic centimetre computer, for 100 
cents). However, this will require fundamental advances in a number of hardware and software areas. 
(C^3 springs to mind as a possible spin, as well as including the "commodity" word somewhere.) 

Currently, smaller inexpensive computers have reduced functionality interfaces (e.g. smaller real 
estate and power/resolution for displays, pen based input instead of keyboards, etc). By contrast, with 
3D projection display, voice in/out, and other interaction modes (gesture, etc), one could consider a 
wholly different style of use. We do not consider a computer for embedded systems work (as in 
phone/PDAs available now), or a pure tablet (a.k.a "network computer"), but a full feature device. The 
best description of the type of interface we envisage, for Science Fiction fans, is probably in the 
1940s novel "Second Foundation", which introduces the Prime Radient, with full fledged computing, 
communications, storage, etc. Such a computer may be 'wearable', although it may well be something 
that one places in large numbers around one's person and property, but also in the broader, public 
environment. Hence surface area might be more significant than cubic volume. Initially, devices may 
come in several parts -- earpiece, eyepiece, pen -- all communicating without wires. As users 
acceptance improves, direct linkage with the senses should become commonplace. One vital piece as 
far as research in computing science is concerned, is that all components must be socketless, to 
reduce deployment costs. No power cord, no video link, no internet socket, no keyboard and no mouse. 
It is these that today contribute significantly to total cost of ownership. And yet the "system" should 
present the full capability of a PC and more. 

The Problem Space

First and foremost, the concept of virtual ownership must be addressed. The move from the 
mainframe to the PC was a move from central to personal ownership. The systems we envisage are a 
move towards fully cooperative virtual ownership. Notions of identity and provenance are 
critical.There are many challenging technical problems here, but also social value: disposable grids; 
affordable computing for the developing world; collaborative filtering; emergency service and disaster 
recovery support; location aware services etc, etc. 

Why the UK?  Because we have ARM, Symbian, the best people in Interaction Design (e.g. the Equator 
EPSRC IRC), and a good systems community, as well as the console games software community, who 
understand fitting a lot of complex systems into small spaces with novel interfaces. In some senses, 
this challenge is to recreate the social and research environment that existed in Xerox PARC that led 
to the creation of the workstation, Ethernet, and WIMP/GUI model of systems, but to the next level 
of integration. The integration is extended through the virtualisation of interaction devices, directly 
through to the human sensorium (sound, light, touch). 

N.B. This is NOT the same as "smart dust" or similar intelligent sensor net computing projects - the 
intention is to retain a full generality of the "PC", but with some shifts in a few ways it interfaces to 
people. Nor is the intention to combine lots of functions in one small device (personal 
communicator/digital assistant/games console/toaster). Although a successful response to the 
challenge would be applicable in those arenas, we would expect special purpose embedded processing 
to be 1000-fold cheaper too by 2012. However, the dynamicity of these systems will be extreme. As 
users move amongst the computing milieu, they will autonomically enlist services as easily as one uses 
physical tools today (knives and forks, whether at home, at a friends or in a restaurant).

The Solution Space

We envisage a component approach to the challenge, with a building block consisting of  a fully 
integrated device with surfaces that can emit and respond to light, sound, touch and RF. The device 
should create an environment around itself, so that it is useful. 

There are Material Science challenges here for the choice and development of appropriate 
silicon/sense interfaces.  Probably a critical CS/EE technical critical challenge will be in the area of 
power, and heat dissipation. 

How will these devices appear from a programmers' point of view as well as an end users'? Will 
C/Java/.NET be used? Will it just be different APIs or might it lead to more radical changes (like 
Smalltalk at PARC)? Ambient, Sentient and Ubiquitous computing are phrases that are starting to 
appear in this context.

However, there are subsidiary challenges in the general architecture of such a device, including Busses 
and storage. 

The most engaging part of the challenge seems to us to be in the area of usability. By dispensing with 
conventional interaction devices completely, we will force the system designer to address the set of 
relevant tasks more imaginatively. Applications such as collaborative document preparation ("word 
processing"), asynchronous human-human communication ("e-mail", "voice mail"), information searching 
and retrieving services ("web browsing"), decision support systems, and so on, may all become very 
different activities.