Google Gears

» see gears.google.com as it lived, on the wayback machine

~ the obituary ~

google gears launched at google developer day on 31 may 2007. it was a browser plugin, meaning you installed it separately, and it added a small set of javascript apis to whatever site you were on: a local sqlite database you could query, a background worker thread for javascript that would not block the main page, a local caching server for offline access to resources, later a geolocation api that used wifi triangulation before native browser geolocation shipped.

the pitch was that web apps needed offline mode. gmail was two years old, google docs was a year old, google reader was less than two years old. all three had the same problem: you opened them, you got a page that only worked with a live network connection. if the wifi dropped, the app stopped. gears was the fix. install the plugin once and any web app that supported it could keep working through the disconnection window.

this worked well enough that a small set of important web apps shipped gears-based offline modes within a year. gmail offline arrived in early 2009. google reader offline shipped earlier. google docs got offline through gears. wordpress had an offline blog editor. myspace of all things shipped offline messaging. really though, the whole idea was that the browser was going to become a real application platform and gears was going to bootstrap it.

i was in my early twenties when gears came out, doing web work between engineering classes in ankara. really though, i installed gears at least twice on my ttnet-connected desktop, once to try gmail offline and once because google reader nagged me about it. the offline part worked. what did not work is that i had a broadband connection at home and reasonably reliable wifi at university, so the offline benefit was small enough that i mostly forgot to sync before disconnecting anyway.

~ the case file ~

~ how it worked ~

gears was a native browser plugin, which is the same category as flash, java applets, activex. it hooked into the browser as an extension and exposed a javascript api that pages could call. the sqlite database ran locally in the user's profile directory and you could do real sql queries against it from javascript. the worker pool let you run background scripts without blocking the ui thread, which was a big deal in a pre-web-workers world.

the localserver module was where the offline magic happened. anyway, you gave gears a list of urls (an html page, a javascript file, a bunch of static assets) and it cached them locally with a manifest. when the user navigated to those urls offline, gears intercepted the request and served the cached version. it was an application cache before the html5 application cache existed.

~ what killed it ~

html5 killed it. between 2008 and 2011 the web standards process figured out that gears was solving real problems in ways that should become browser-native, not plugin-mediated. localStorage handled key-value client-side storage. web sql (then indexeddb) handled structured local storage. web workers handled background scripts. the application cache spec handled offline resource caching. the html5 geolocation api replaced gears geolocation.

in november 2009 google published a blog post saying it was moving its offline efforts from gears to html5. development on gears officially stopped on 19 february 2010. chrome 12 (march 2011) removed the built-in gears plugin. gmail and google calendar dropped gears support on 1 december 2011. the offline gmail app was ported to a chrome extension that used html5 apis directly. the transition took about two years of gears being the recommended path becoming gears being explicitly deprecated.

~ what got buried ~

the specific plugin, mostly. the concepts that gears prototyped all survived, in most cases as much cleaner html5 apis. offline gmail still exists. google docs offline still works. what disappeared was the model of shipping a browser plugin to add features that would eventually become platform standards. gears was one of the last successful examples of that pattern (silverlight and flash were the others, both also now dead). after gears, the browser vendors treated the platform as complete-enough that plugins were no longer the way to add capabilities.

the second thing worth noticing is that gears was a rare case of a plugin doing exactly what it promised and then getting deprecated in a way that respected users. the gmail offline transition to html5 was reasonably clean, offline google docs kept working through the migration, the applications that had used gears mostly got equivalent html5 replacements. it was a plugin retirement that did not orphan the users who had actually depended on it, which is not the pattern most plugin retirements followed.

@}-->--

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

« previous  ·  random  ·  next »