Posts on this blog represent my opinion. It may be my considered opinion on the basis of my formal study of law and technology. But it is not legal advice. It must not be treated as, or acted upon as, legal advice and no liability is accepted for doing so.

Tuesday, 14 April 2009

Amazon, Twitter, and the Gay Books Purge that Wasn't

A little bit of background: over the holiday weekend, news began to spread that Amazon had done something rather odd and disturbing with many, if not most, books that dealt with lesbian, gay, bisexual or transsexual themes. They hadn't been removed from sale, but their sales rank had been suppressed. As well as being a direct indicator of a book's popularity, this is a key factor in Amazon's automated book recommendation system, so a book without a sales rank is far less likely to be offered up to potential customers. I know a few authors, and their Amazon sales rankings are something they take a keen interest in. For Amazon to remove them for a whole category of books is naturally going to concern both the authors of those books and anyone interested in that category. And when that category is LGBT books... well, you can imagine that suspicions of Agendas, or Moral Panic, or even Censorship started to circulate.

And circulate they did, thanks to Twitter. April 2009 might well go down as the month that Twitter went mainstream, firstly with The Guardian's April Fool, and then with the Twitter '#AmazonFail' tag, by which Twitter users alerted to the issue could chose to follow comments and updates about it.

(Two points here. Firstly, this highlighted for many people the power of Twitter tagging as showing a use for what is often regarded as a rather trivial medium. A Twitter tag in effect allows users to get an instant paging service on a current topic in short, bize-size form that can easily be pushed to a phone or PDA. It's an excellent way of forming an instant community-of-interest, as this example showed. Secondly, I wonder if anyone will do a study into the way that 2009 has seen '[name]gate' as the label for a scandal be supplemented by '[issue]fail' as the instant term for a controversy, at least online?)

I won't dive into discussion or explanation as it has been done far better elsewhere. US editor and blogger Patrick Nielsen Hayden has made some sensible observations (and there are a few more in the comments to that post, admittedly amidst a lot of wibble). The Seattle Post-Intelligencer blog has what seems to be credible news from sources within Amazon on what happened. But most interesting to me are the comments from tech blogger (and long-term friend) Simon Bisson on what this tells us about Amazon's infrastructure.

And what it tells us may not be good news for Amazon, or indeed its shareholders. To quote Simon:

"The simple answer is Amazon's architecture. It's highly distributed, and there's no operations team. Each component (and over 200 go into a single page) is run by its development team, of four to five people. They are responsible for its features, its development - and for making sure it runs effectively."

We had a term for this when I was in the RAF: "Spring-Loaded", as in 'crammed full of cogs and springs that will explode in a shower of little bits of brass unless the lid is screwed down very tight'. Another term is "System of systems", popular as a cool-sounding buzzword but a phrase that should strike fear into the heart of anyone who understands that ten fragile things stacked up together are in fact likely to be more, not less, delicate than one fragile thing alone.

I don't think for a second this was an evil reactionary plot by Amazon to purge itself of LGBT publications or to appease the Religious Right. Whilst that, if true, would have been very bad for Amazon's reputation, I think the actual explanation may in the long run be even worse. If it turns out that such an embarrassing incident could have arisen from a single coding error, and that Amazon's infrastructure allowed the error to pass undetected, propagate around the world and then take days to fix, then it rather makes the world's best-known online ordering brand look like a massive house of cards. At the very least, it will be an object lesson both in scalability of architectures and in corporate image management in the age of Twitter.

I look forward to the e-Commerce conference papers with interest.

1 comment:

marypcb said...

*every* online service has proven the fragility of its architecture by a massive failure of some kind, from AOL onwards (Gmail and Magnolia being the most recent). What's really noteworthy about Amazon is that their policy is to get good code by making programmers admin their own live code; don't like being woken at 3am? write better code. My phrase for this approach is 'accident waiting to happen' - development, testing, deployment and admin are no more overlapping skillsets than ability with a soldering iron and driver design are for hardware...