Wired on Friday : At first glance, Google's new map feature (http://maps.google.com/) doesn't look that much different from any of the other mapping services online. The feature set is broadly the same: you type in an address and get a map; you can zoom around to see neighbouring features, and you can get driving directions.

As is usual from American companies, the maps are for the United States only. The data is drawn from the same few sources as other online map services. Nothing special here.

It seems a little disappointing that Google should be attempting to enter this area. Yahoo and Mapquest both have established map services. They're perfectly reasonable. Google uses them itself. When you type a US address into the Google search box, it will often offer you a link to one of these sites, pre-configured to show the address.

Before using it, then, this new service seems to be part of the overreach that most IPOed dotcom companies fall victim to. It certainly isn't an improvement in what most people know Google for: searching the Web. It's a dilution of Google's core competency.

Or is it? Two or three clicks in, and you notice a profound difference between Google's offering and those of other companies. There's none of that stop-start, hit button, wait, watch as the browser window slowly reloads with your results.

When you drag the zoom lever, the map grows without disturbing the surrounding text.

When you drag the mouse over the map, it scrolls almost seamlessly, giving the feeling of a tiny window over an entire United States map. Request driving directions, and the route is redrawn over your existing map as you watch. It feels, in other words, as responsive and fluid as a desktop application, not a web application.

Google Maps is an example of a new kind of web page; one that doesn't require constant reloading to refresh its contents, but which constantly, dynamically, redraws itself. While the famous search engine has pioneered this approach on Google Maps, its email service Google Mail, and another beta experiment, Google Suggest (which throws up popular search terms as you type), it is certainly not the only entity taking this approach.

Amazon's competitor to Google, A9, has similar features. The airline search service, SideStep, updates air flights as you watch. Other design companies, including San Francisco's web consultancy Adaptive Path, are working with clients to include the technique in their own websites.

Jesse James Garrett, the founder of Adaptive Path, calls the approach Ajax, which describes the two technical components that lets the system work: asynchronous Javascript and XML. Garrett says the main challenge to Ajax isn't technical, it is the challenge of imagining how it might be used.

The technology, at least on the browser side, shouldn't be challenging. They've been around for years. Ajax pages work with Internet Explorer 6, which was first launched four years ago. So why haven't we seen such pages before online?

Has the imagination of the Web been so lacking for so long?

Some of the reason why the Web has not seen many pages with this degree of interactivity is perhaps less to do with the limitations of the browser, but the limits of web servers, and the understandably risk-averse people who operate them.

When websites can have a few hundred users one day, and millions the next, being able to predictably scale up your page delivery as your traffic increases is an important consideration.

The worst thing that could happen to a website company is for their site to "fall over" - break so irretrievably as to refuse new visitors - on the very day of their greatest popularity.

To concerned engineers, a page which reloads its data is something of a challenge. It's not something that you would encourage your web page designers to pursue without caution.

One company that realized this, several years ago, was a Silicon Valley start-up called KnowNow. Composed of very smart web theorists, the company worked hard on building servers that could cope with a more dynamic Web, and scale predictably.

Admittedly, the dynamic Web that KnowNow was attempting to engineer was a little more complex than Ajax. "The technical difficulties with Ajax are equivalent to coping with multiple reloads of the same page," says the company's founder, Rohit Khare, "But it's just the first step to a more interactive, asynchronous Web and there are definitely technical challenges to creating that new Web." Perhaps the primary reason we haven't seen this before is because so few companies were willing - or had the research and development budget - to learn more about Ajax-style pages.

Now Google is raising the bar again. Once you've played around with the company and its fellow Ajax-users' dynamic experiments, you see how it could improve almost every web page.

Google's core competency, it is becoming increasingly apparent, is not so much search, it is its institutional knowledge of the possibilities of the Web: both on the browser side and the server engineering required to scale up such new services to a reasonable size.

It's an odd economy of scale: Google isn't able to do these things because it is bigger, but because it's learnt how to grow bigger without - so far - breaking. Anyone can put up a web page - or a mapping service - Google is saying, but how many can make it dance like this, and never fall over?