Google Web APIs

~ the last words ~

In 2002 I was eighteen and the internet I had was a cafe in Turkey, which meant Counter-Strike and whatever would finish downloading before my hour ran out. Google had just published a way for programs to ask it questions, and you needed to sign up for a key, and nobody in that room was writing anything that needed one. This is a grave for a thing I only understood the importance of much later.

Google Web APIs arrived in April 2002 as a beta. It was a WSDL file and a SOAP endpoint, which in 2002 was the respectable way to build a machine interface, and it answered a search call plus two smaller ones, a fetch of Google's cached copy of a page and a spelling suggestion. Nelson Minar wrote it. You downloaded a developer kit, registered for a key, and your program could now see the index.

The economics were the memorable part. The key was free and tied to a person, capped at five hundred queries a day and later lifted to a thousand. Commercial use was not allowed. For a few years that cap was the ceiling on an entire genre of software: if your idea needed more than a thousand Google queries a day, your idea was not going to exist.

On 5 December 2006 Google stopped handing out new keys and said it was no longer actively supporting the service, which it carried on supporting anyway for nearly three more years. The actual end came in a post dated 31 August 2009 titled as a retirement rather than a shutdown, and even that slipped by a week. The sentence in it was "we'll be shutting it down on September 7th instead."

~ the record ~

~ what people built with it ~

The obvious thing was the Google box, a small widget you dropped on your own web page that ran a live query and showed the results, which is how a lot of personal sites in 2003 proved they were current. Underneath that sat the rank checkers, the tools that watched where a given page sat for a given term week after week, which is most of the early search-marketing industry in one API call.

The less obvious thing was research. A thousand queries a day is enough to measure something, and for a few years academics studying the shape of the web were doing it through this key, legally, with Google's blessing. When the key went away that work had to go back to scraping.

~ the slow retirement ~

The three years between the deprecation and the shutdown are the interesting part of the file. Google's own position in December 2006 was that it was not planning to switch the service off for anyone already using it, and it kept that promise longer than most companies do. Existing keys worked. Nothing broke. New developers simply could not get in.

That is a specific kind of death, where the product is sealed rather than killed, and the population using it can only shrink. By 2009 whoever was left had been grandfathered for so long that the shutdown post could reasonably call it a retirement.

~ the argument it started ~

Leonard Richardson wrote the sharpest response on 20 December 2006, two weeks after the deprecation. His point was that the replacement, the AJAX Search API, was the better piece of engineering and the worse deal. SOAP was heavy and the new thing spoke plain HTTP, which everybody agreed was an improvement. But the new terms, as he put it, meant "you're only allowed to use it through one library in one language in one environment for one purpose."

That trade kept repeating for the next twenty years across every platform that opened an API and then closed it down to one approved client. The SOAP envelope was the thing that looked dated in 2006, and the license was the thing that actually mattered.

@}-->--

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

« previous  ·  random  ·  next »