Site icon IT World Canada

What Amazon’s catalogue glitch says about cloud-based metadata

Software glitches used to simply cause headaches. Now they trigger conspiracy theories.

The cataloguing error that plagued Amazon.com over the weekend was in all likelihood not a malicious attempt to censor books written by gay and *** authors, though several online commentators have suggested otherwise. For all Gore Vidal’s huffing and puffing, there were still titles from Bret Easton Ellis and Chuck Palahniuk. Still, even a headline for an Associated Press story about the incident put quotation marks around the word glitch, which implies the error might not have been all that accidental. (A hacker has claimed credit for what happened, but if that were true there seems little reason for Amazon to take the heat.) Amazon’s prompt reaction to the problem and its willingness to admit its scope is enough to convince me that the company is not inherently homophobic. If there’s any phobia apparent from this incident, it’s the fear of not having access to cloud-based data.

What’s remarkable is that, to the best of my knowledge, this glitch didn’t necessarily affect the delivery of existing orders. No one’s personal information was compromised or lost. This was, for most people, a search engine problem, but the reaction would have been more appropriate to banking customers who weren’t able to use their debit cards.

Amazon is not only one of the leading success stories in e-commerce, it’s been an early champion of cloud computing with its EC2 and S3 services. From an IT perspective, there doesn’t appear to be anything related to the functioning of its data centres that were related to the cause of this glitch. This was a content management problem, and it underscores the lesson that if information isn’t where it’s supposed to be, it doesn’t matter whether your applications are on premise or hosted in the virtual beyond.

In this case, the issue was not merely that data disappeared. It’s that such a particular kind of data disappeared, so that the glitch looked discriminatory. Yet the whole point of content management is to slice metadata further and further until it becomes more effective for search engine optimization, targeted sales and other aspects of “long tail” business approaches. If Amazon’s catalogue treated all books the same, this error might have seemed less suspect, but Amazon would also probably be less successful as an online bookseller.

As more data gets put into the cloud and offered out in customer-facing applications, companies may find themselves in the uncomfortable position of explaining how and why they organize information for the most personalized customer experience possible. Catalogues and other repositories are supposed to discriminate. When they work properly we call them searchable. When they don’t, we call it censorship.

Exit mobile version