How do we know what they do when they are mobile?

Ericsson position paper at the W3C Web Characterisation workshop

Author: Johan Hjelm, W3C/Ericsson

Ericsson is a data communications infrastructure company. This means that we must understand the customers of those who buy the equipment we produce - the end users.

But the mobile environment poses some problems in characterisation. How do you know users are mobile, when they come in like they were on modem lines, using the same PPP connections and TCP/IP stacks as fixed-line users? This is something we need to investigate, to be able to recommend areas where we can optimise services for them.

We are also interested in modelling tools to be able to produce better networks and systems. If we can gain understanding on how users will behave, not only in the present-day networks, but in coming generations of systems, we will be able to serve them better. This is true both for individual behaviour and their behaviour as aggregates. Knowing how systems will be loaded enables us to build network infrastructure equipment that is optimised for its task, even in a rapidly changing environment.

This is especially true in the mobile field. Currently, the capacity of the bearer is limited, in some cases to as little as 22 byte per second. However, this is set to change. With the 3rd generation technologies (UMTS/IMT-2000), users will get a capacity of 2 Mbps. Another difference from today is that networks will be packet-oriented, rather than circuit-switched, as today. This can mean drastic differences in the capacity available for the user, and significant changes in latencies and other aspects of network characteristics.

As seen in the chart below, this migration is already taking place.

This rapid pace of deployment will mean that the networks will be deployed at the same time as we want to find out the users behaviour. This delicate chicken-and-egg problem can be solved by working intensely with developers and early adopters, setting up test services before commercial deployment, and conduct laboratory-scale experiments with users, to enable us to obtain the usage patterns which will be propable in the networks. How to do this is not self-evident, and will need to be worked out in collaboration between the equipment manufacturers, the network operators, and the Web Characterisation Activity.

The deployment of new technologies does not mean that the older technologies will go away. The frequency ranges allocated to older services will not be retracted, and as the equipment is written off, service prices will propably continue to drop. This will mean an increasing pressure on the operators of new systems to bring new services to market, as they strive to keep their services perceptibly different from those of their competitors (this is something that is already taking place). With the efforts of the WAP Forum, these can be expected to start occuring during the next year. Does this mean that user behaviour will change? We do not know, but the methods outlined above can be used to find out.

Devices will also keep changing. We expect they will continue to exhibit a reverse of Moores Law: instead of prices being fixed and capacity expanding as in the PC industry, showing prices dropping and capacity being fixed in the mobile devices (although there will, most likely, be a leap over to a new generation of devices within the next few years, so called "smartphones").

The form factor of those will also continue to change, making possible changes in user behaviour. Anecdotal evidence from the Scandinavian countries, where mobile phone usage is the highest in the world, already shows the behaviour in using telephones changing. And statistics clearly demonstrate that people, especially young people, do not subscribe to fixed line services, choosing instead to go entirely mobile. There is no question but that these devices, when bandwidth becomes accessible, will be used for broadband applications such as video.

However, we know very little about what should be regarded as normal user behaviour. This of course makes it impossible to discover what the effects of the changing communications environment is. We do believe that to begin, we have to create a picture of the normal usage patterns. In the wireless environment, the use is so widespread in certain communities that it should be possible to chart it. From these behaviour patterns it may then be possible to rationalise changes derived from the changing communications environment.

We realise this is a complex and difficult task, as the user behaviour is very hard to demonstrate before the technology is deployed. Still, we find the effort worthwile, and will support it with 30 percent of Johan Hjelms time.