Internet business 101

By G.F. | SEATTLE

THE latest newcomer to the e-lending business was the first to be called out. Like other similar sites, Lendle relies on loans of books purchased for Amazon's Kindle reader. But on Monday morning Amazon turned off the content tap that feeds the site; it was not switched on again until Tuesday evening. The brief outage demonstrates a fundamental truth about the internet: if you don't own the data you need to run your business, you're dependent on the policies—and whims—of the parties that do. Jeff Croft, Lendle's founder, says he will not make that mistake again.

Amazon resorted to a clever ruse when blocking Lendle. No threats of legal intervention were involved. Instead, Mr Croft says he received an email from the Amazon affiliates program saying Lendle had violated the terms and had been locked out from a database of content and deprived of the share of sales of items linked to from its site. Five of Lendle's competitors appear not to have been similarly afflicted. (Amazon did not respond to a request for a broader comment.)

In a letter to Mr Croft, Amazon explains that Lendle retrieved content from Kindle owners' library pages and used information obtained through an Amazon database intended for sales affiliates for purposes other than selling merchandise. Sadly, both statements appear to be true. It did not help that Lendle links to no other booksellers and creates sales for Amazon.

Amazon has long proscribed scraping, as the programmic retrieval and sifting of web pages intended for viewing in browsers is known. Programmers employ scraping to gain access to information ordinarily off limits to automatic processing. In this case, the method allowed Lendle users—with a single click—to fill their accounts with all the Kindle books they owned that were available for borrowing. The feature gave Lendle a leg up over other e-lenders. It also almost proved to be its undoing.

Amazon prefers—nay, demands—that partners make use of the massive array of information about items for sale that the firm provides only through an application programming interface (API). An API acts like the circulation desk at a library with locked stacks. A programmer may approach the virtual desk, and with the right identification and invocation, receive materials for use on his site. A programmer need only cross the librarian once to find his feet on the street, with the collection out of reach. API servers are typically managed separately from public web sites, preventing overuse of such data from bogging down sales. (This Babbage has worked with Amazon's API for years and has had no problems thanks to studious adherence to the program's rules, even as they have grown increasingly baroque.)

Amazon requires that information it provides from its product-advertising API (as it terms this data) be used solely to promote sales. Each use of a datum must be connected to an affiliate link back to the appropriate item or section at Amazon.com: "all uses of this content must serve the principal purpose of driving sales of products and services on the Amazon site," the letter warned, suggesting that Lendle fell foul of the rule. Lendle did use some data to promote lending separate from purchases, but Amazon's rescission presumably means the powers that be changed their mind and decided that Lendle did, in fact, comply.

This kind of stumble hardly comes as a surprise. Since Google popularised the notion of information mash-ups, where several kinds of live and archived data are combined to produce new value, companies have come to see information as a freely available resource, draconian terms of access notwithstanding. Google tends to be rather liberal in this regard, but does put some limits on daily usage. It also requires payments for commercial, or particularly heavy, deployment of its services, as with maps.

Developers of apps for Twitter were recently given a start when Ryan Sarver, the microblogging platform's top API man, said the firm would only give certain kinds of new desktop and mobile software access to its content. Could programmers build client apps that use Twitter's databases but do little more than mimic what Twitter itself does? "The answer is no," Mr Sarver wrote categorically. The company's desire to plug its own software is understandable, but the move adds uncertainty to any company or programmer who might consider writing a fee-based or ad-sporting app. Apple is far more consistent now than two years ago, but it is still impossible to know in advance whether a newly created iPhone app mixes and matches Apple features or extant third-party programs in admissible ways. (A colleague's sutra parody was denied by Apple because there were too many Kama Sutra apps already—as if that were possible.)

In a note on his website posted on Tuesday, after Amazon had restored access to its database, Mr Croft wrote that he and his team "have come to realise we need to work towards a Lendle product that does not rely on APIs provided by Amazon or any other third party." Lendle has learned its lesson.