Netflix Public API

» see developer.netflix.com as it lived, on the wayback machine

~ the obit ~

Netflix opened its public API on 1 October 2008, back when the company still mailed red envelopes for most of its revenue and streaming was the free extra on top. Any developer could register for a key and read the catalog. You could check whether a title was available and, with a member's permission, read and reorder their queue.

The internal name for the thinking was the thousand flowers model. Netflix would expose what it knew and let strangers plant whatever they wanted in it, on the theory that some of those plants would turn into customer acquisition the company had not paid for. That was a reasonable bet in 2008 against a subscriber base in the single-digit millions and a competitor still operating physical stores.

By the time it closed, Netflix had around thirty-five million subscribers in the United States and another twelve and a half million abroad, and Daniel Jacobson, the engineering director who ran the API, had written publicly that the garden had mostly not grown. Traffic through the public API was a rounding error next to the traffic from Netflix's own apps on televisions and consoles, and the public API's design was actively in the way of serving those devices well.

Netflix announced the closure on 13 June 2014, stopped issuing keys that same day, and switched the endpoints off on 14 November. A handful of partners kept access. Everyone else started getting 404s.

~ the rap sheet ~

~ what people had built on it ~

Instant Watcher was the one people actually mourned, a site that existed purely to tell you what had just arrived on Netflix streaming and what was about to leave, which is a question Netflix itself has never answered well in its own interface. Expiry dates in particular were something the API surfaced and the app never did.

The rest was a long tail of queue managers and small apps that answered one question, usually whether the thing you had just read about was streamable tonight. Most of them had already drifted away by 2014, because the API had been frozen for years and the catalog had moved to streaming while the data model stayed shaped around discs.

~ the argument it settled ~

Leigh Dodds wrote the clearest response that June. His point was that an open API is a bet on uncertainty, and once a company knows exactly what its product is, the bet has no payoff left. Netflix in 2008 did not know what it was going to be. Netflix in 2014 knew precisely, and an open API is expensive insurance against a question you have already answered.

What followed was the standard shape. Netflix went on building internal APIs at enormous scale and went on talking about them at every conference that asked, so the engineering stayed in public view while the door quietly closed behind it. By 2014 the industry had learned that an open API is a growth tactic with an expiry date on it, and Netflix shutting one of the most-cited examples made that lesson hard to argue with.

~ the distance ~

Netflix did not open in Turkey until January 2016, over a year after the public API was already cold. So the entire six-year life of this thing ran in a service I could not legally subscribe to from where I live. I knew the company as a name in American tech coverage and as the reason people argued about DVDs by mail, which is a strange way to know a company that would later be on my television every night.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »