Wired on Friday: What does Amazon do, exactly? Back in 1994, the answer would have been simple: Amazon sells books. A few years later, one would have to broaden the answer to include Amazon's sales of a battery of other products, from lawnmowers to dresses. Then matters would grow blurry: Amazon "z-shops" appeared, with small sellers camping out on Amazon's pages in return for a cut of the proceeds.

Then Amazon launched its own search engine, A9.com. Then it introduced a map service. What do free web map services have to do with books, apart from, perhaps, competing with atlas sales? And what, in particular, has online storage to do with Amazon - and why exactly have they announced a way for strangers to save gigabytes of personal data on their ostensibly book-selling webservers?

There is another way of seeing Amazon, and that is as a technology company. There was a time when the company itself promoted this viewpoint - largely, one guesses, because tech companies survived the 2001 tech market meltdown rather better than online retailers did.

But, despite the description, most technologies that Amazon offered around its retail core business were, in reality, just ways of getting more people to visit their shops.

The earliest and most public example of these "technologies" is Amazon's associate programmes.

Hundreds of thousands of websites across the internet offer links to particular books or products on Amazon, almost incidentally to their main content.

A blogger, for instance, may mention that he has enjoyed re-reading Ian Fleming's Moonraker, and provide a link to the book's page on Amazon.

Wandering surfers who impulsively decide that they too must learn about Mr Bond's fight with Sir Hugo Drax click on the link and buy the book. The blogger gets 10 per cent of the sale.

For most, Amazon associates provide nothing more than beer money. For others, however, redirecting visitors to Amazon provides a lucrative income. For these external sites, Amazon really is providing new technology. It offers several "web services" for its serious associates.

These programmers' gateways into the inner world of Amazon's transaction processing backend allows these power sellers pour over and maximize their profits, while binding them more and more tightly to Amazon. A humble book seller could switch from Amazon to, say, eBay easily. A seller whose high profits depend on them finding out, say, what books are selling for the highest second-hand price, would be tied forever to the code written especially to discover this on Amazon.

Occasionally, though, Amazon really does offer technology utterly unrelated to its retail pursuits.

A few years ago, for instance, Amazon announced that it was providing a web service called "Simple Queue Service".

And what did this do? Well, it did one thing well - it let you pile up data in queues on Amazon's servers. You could send a blob of data to Amazon, send another blob, and then retrieve them in the same order. Fascinating, no? Well, yes, in fact. Writing a system to reliably queue and provide data is one of the problems in designing large scale e-commerce projects that is so simple-seeming and ubiquitous that every coder is obliged to re-invent it.

Smart coders, however, know that the devil is in the details and that some of the most intractable and confusing problems in building a system that can grow from 10 to millions of users arises from a badly designed queuing system.

Enter, then, Amazon. Amazon, as one of the largest e-commerce systems in the world, had clearly solved the queuing problem to its own internal satisfaction. This meant that they had also solved it for almost any other e-commerce site on the internet.

However, only the very largest systems' queuing requirements would even make a blip on the amount of processing that took place on Amazon's internal servers - so why not offer them a chance to use the same queuing technology?

The company could charge a tiny amount, and not only benefit from reselling its computer systems' power, but also tie other companies, once again, to Amazon's network.

The Simple Queue Service is still in beta and Amazon is not charging for it, even after two years of operation.

But the company launched a similar, and even simpler web service this week. For 15 cents a gigabyte per month, anyone can now store data of any kind of Amazon's servers (uploading or downloading the data costs 20 cents a gigabyte).

The new web service is called "S3". Once again, Amazon is attempting to offer (and profit from) the work it has been obliged to conduct to be a 24/7 online store.

Data and records cannot go missing in Amazon's databases and so, the promise goes, your data will be similarly backed up and protected.

Amazon's offer is competitive with dedicated commercial storage services, and there's been plenty of interest in S3, largely because of Amazon's brand notability. But perhaps for SQS and S3, the looming nature of Amazon is a little too distracting.

Buyers and sellers who tap into Amazon's associate web services know that they're clinging close to the online retail giant. But trusting Amazon to sell your books is rather different than trusting it with your data, or even the core of your business logic.

While it may be hard to work out exactly what Amazon's core competency is these days, it's not difficult to see third-party storage and processing as a side-line. And if there's one thing stores are good at, it's dropping sidelines when they're unprofitable - no matter how precious or vital they are to their ex-customers.