Amazon has four customer sets:
? Buyers
? Sellers
? Web Site Owners (Associates)
? Developers

Amazon has made a considerable investment. $300 million in distribution. A staggering$900 million in technology. And about $700 million in marketing.

Amazon started to go down the Web Services road when Tim O'Reilly wanted sales dataabout the O'Reilly books, and was getting sick of screen-scraping. He wanted thisdata as XML. Stealth projects were already underway, and the Amazon Web Services effortwas born.

AWS offers both a SOAP and a REST interface. They also have two different models forreturning data: the "lite" schema and the "heavy"" schema. The "lite" schema justreturns a simple set of product information -- the "heavy" schema returns practicallyeverything& including customer reviews, sales ranks, "browse nodes" that indicatewhere that product lives in the Amazon information space, and just about anythingelse you can think of. The amount of information that Amazon keeps about a specificproduct is pretty impressive.

The NBA Store (http://store.nba.com) is a custom store that uses Amazon web servicesunder the hood. NBA.COM uses AWS 3.0's Remote Shopping Cart feature, that allows externalclients to securely store Shopping Cart information in Amazon's database. When thecustomer wants to check out, they get transitioned to a co-branded checkout site that'shosted solely on Amazon.com.

Jeff is talking now about how Amazon has implemented a publicly consumable XSLT transformationengine. This is used to allow small business to use Amazon.com web service enginefrom a 3rd party hosting service that doesn't support advanced things like web servicecalls. The seller can build a static HTML page with links to Amazon.com product URL's.However, in the URL passed to Amazon lives a reference to a seller-provided XSLT stylesheetthat Amazon has cached. When Amazon processes this request, it takes the XML productinformation returned from its own web services, runs the XSLT stylesheet to createseller-specific co-branded HTML, and returns that whole chunk of HTML back to theshopper the originated the request.

Inside of AWS, the web services are aggregating information from a number of internalAmazon systems (pricing, catalog, sales). Any one of these servers might fail duringthe processing of a request. Instead of failing the entire request, the AWS systemwill return a partial response, and the response will contain metadata as to whatcomponents of the response are missing.

Amazon URL builder:
http://www.syndic8.com/~jeff/xml.html

Interesting note: about 80% of the requests that Amazon receives are processed viathe REST interface.

Anybody who wants to see a real-world example of the business value of web servicesjust needs to look to Amazon. The type of things that they are doing with web servicesare exactly the type of things that web services were designed for, and their makingmoney doing it.