Microsoft is currently using a short video to illustrate the concepts and benefits of its .NET web services initiative. For those of you who haven't seen it yet, it tells the story of Steve, a business guy, who has an unfortunate penchant for standing in the way of cycle couriers.

The luckless Steve, on the occasion we see him, gets knocked down in San Francisco, which is not his home town. Using his souped-up personal digital assistant (PDA) he finds the nearest doctor. What follows is an example of how his personal financial (insurance, medical plan) and medical information details get transferred at blinding and efficient speed over the network, with intervention being possible by him in real-time, while he spoke to the receptionist, allowing him to be efficiently treated before being sent on his way - alas, perhaps not surprisingly, before being waylaid by another two-wheeled assailant.

Microsoft staffers who show this video say it illustrates what life will be like in the coming .NET golden age.

In May this year an article appeared in Scientific American. In it, a guy called Pete helps his sister Lucy out when their Mom needs some medical treatment. Using his handheld browser, Pete accesses a range of financial and medical records in order to progress his mother's treatment, while being able to ask questions about location and cost of possible therapists in real-time - I think you get the picture.

What's going on here? We're very far from accusing Microsoft of having stolen the scenario from the authors of that article. But when you understand that the writers of the latter included no less a figure than Tim Berners-Lee, 'father' of the World Wide Web, you can begin to understand what's going on.

Whatever Berners-Lee did 'next' was always going to be of massive interest. And when that turns out to be a hugely ambitious plan to make the web a smarter place, it's forgivable to see why someone at Redmond might have been interested in that article. One take on Berners-Lee's project might well be the kind of services Microsoft is trying to articulate with .NET and Sun Microsystems with its rival ONE framework. But some observers think it's a far more ambitious - maybe even pernicious - plan than either.

Berners-Lee's goal is the creation of what he calls a 'semantic web'. By this Berners-Lee and his team are referring to giving intelligent agent software the ability to understand the meaning (semantics) of the information they are roaming over in order to make their searches more inherently meaningful and efficient.

'When it comes to data, we live in a pre-web era,' said Berners-Lee in a recent speech. 'We spend huge amounts of money on employing integrating consultants to analyse the data of one company and bring it to another. Here lies a possibility for the web. It has the information. There are clear mathematical algorithms that can define the data. If we can design a language to read this, then there can be communication. Being able to access data which computers can help us analyse may have dramatic consequences in the long run.'

And that's a good thing, thinks Jacques Hal‚, director of research for UK analyst group, Butler.

'Berners-Lee isn't the only one doing this kind of work, but he is the one with the clearest vision,' says Hal‚, who has been working in AI (artificial intelligence) and other advanced topics of computer research all his career. 'It's becoming more and more apparent that we've hit a kind of semantic glass ceiling, and that we can't get much further using Boolean search queries.'

But how to take the next step is a huge problem, says Hal‚. 'Using today's web you cannot get a satisfactory answer to a very simple, mundane question like "I want to buy a yellow car in Boston". The computer - or rather, the web search engine - knows about Boston, Massachusetts and what a Honda is, maybe, but not what colours are or how to put that together. It turns out that's a very complex problem involving more than one answering technique. This is where you start entering the realms of using networks of queries, or networks of cooperating agents, turning the resources of the internet into a distributed search engine.'

For Hal‚, recent announcements on the concept of 'grid' computing - creating large ad hoc networks of computers via the web that share resources - will be the infrastructure part of such an intelligent future net, while work like Berners-Lee's represents the first articulation of the intellectual framework it will need.

But don't forget ontology, and its problems. 'Ontology is a branch of AI that is crucial here - it's how we as humans classify the world, how we know gold is a type of metal, then a type of mineral, then something precious, and also a colour. It's very, very hard to do this right, though. We're not there yet, and that is the main issue for work like this. This is a very long-term project, perhaps as much as ten years off. Web services like .NET are possible applications, but we need to sort out the problem of metadata first.'

Observers like Hal‚ think the semantic web is to be welcomed. But as we indicated at the start, not everyone thinks so.

Tom Welsh is an independent software consultant with over 20 years in software development, mainly at enterprise computing level at what was DEC, now part of Compaq. 'I am deeply suspicious of any plans to let computers run our lives, as even the most basic programs are very hard to do right. There is so much scope for things going wrong here that it could be little short of disastrous if this got implemented.'

Welsh also cites ontology as the key blockage. 'We all know about the idea of the thesasurus but none of us agree as humans on the taxonomy of the "real world" and I can't see how we can expect to teach computers to do it any better.'

And then there are the sinister aspects. 'Berners-Lee has always worked in the supportive context of a closed community. The web itself was originally defined as a kind of library where researchers could browse each other's works-in-progress. It was also very loosely coupled: when you ran into a problem you just gave up and put up a 404 message. This is completely different and will need to be much more closely coupled. Plus, the web is now a commercial operation, one used for propaganda, marketing, and ways to make loads of money.

'Commercial considerations could have a hell of an impact on the sorts of ontologies that end up being used. And if programmers from big companies can't agree on semantics, what of the ordinary person's point of view?'

Welsh therefore urges extreme caution. 'I hate the examples of rushed yuppies always needing immediate information. When I need information on insurance I ring up my financial adviser, not ask the web. This is doubtful value for real people, and for any new technology to cross the Geoffrey Moore "chasm", it needs to exhibit real benefit with no downside. Irrational exuberance is needed to prime the pump but it might take 25 years to work through the issues we're talking about here.'

Two contradictory views, then. For Hal‚, the future is clear: 'Information managers should be tracking developments like the semantic web, but not expecting products for a long time yet. But we will get there, and in the meantime prepare for it by deepening your knowledge of the metadata you need to define.' For Welsh, the semantic web is not something that will ever fly: 'We will have to get software we trust, and that is nearly as intelligent as we are, for it to be realistic. That will take a few years yet.'

The problem is that both are right. Because, like it or not, web services will need something like the classification underpinning of something like Berners-Lee; we do urgently need a smarter web; and at some point soon your MD will see that Microsoft video and ask when he'll have something similar.