OpenSocial

» see opensocial.org as it lived, on the wayback machine

~ the goodbye ~

opensocial was a spec, not a product. i was on orkut as a consumer in the '06-'09 window with the rest of the turkish university crowd, and orkut was one of the first sites to ship opensocial gadgets, but i never wrote one. the developer audience for the spec was the same audience facebook's f8 platform had already captured. i remember the opensocial announcement on hacker news in november '07 and the immediate counter-positioning by facebook fans who treated it as a defensive play from a company that had already lost the platform race.

the spec itself was reasonable. google's david glazer led the architecture, working from the experience of running orkut's own internal gadget system. opensocial defined a common javascript api for the things a social-network app needed: read your profile, read your friends, post to your activity stream, fetch a list of friends' recent activities. an app written against the opensocial api could in principle run on any container that implemented the spec. the founding adopters that day were google orkut, friendster, hi5, myspace, linkedin, salesforce, ning. together they claimed a user base larger than facebook's at the time of launch.

the user-base numbers were the strongest argument the press made for opensocial in its first weeks. facebook in november '07 had maybe 50 million active users. the opensocial coalition added up to several hundred million across all the member networks combined: orkut on the brazilian-and-indian side, myspace's still-dominant us numbers, hi5 in latin america, friendster in southeast asia. on paper the spec had distribution. in practice the network with the developer mindshare was the one that mattered, and the developer mindshare had already gone to facebook.

the practical problem appeared inside the first week of launch. security researchers found that several of the founding members had implemented the spec with holes that let one gadget read another gadget's data on the same container. the press story was that opensocial gadgets had been demonstrated as exploitable within twenty minutes of launch, which was not entirely fair to the spec itself but did capture the running theme of opensocial's deployment: every container implemented it slightly differently, with slightly different security and capability profiles, which broke the write-once-run-anywhere promise from day one.

~ the chronicle ~

~ the breakout ~

the opensocial foundation incorporated in march '08 as the formal home for the spec, with a board that included google, myspace, yahoo, and the founding adopters. opensocial 0.8 in may '08 added the activitystreams protocol, which let gadgets post structured "user did x" events to a network-wide activity feed. opensocial 0.9 in april '09 cleaned up the rest api story and added the gadgets-as-iframes deployment model that became the default for most containers. the foundation kept iterating the spec through 2010, when the deployment side was clearly losing.

the part of opensocial that actually shipped is mostly in the activitystreams 1.0 spec that came out of the same foundation working group. activitystreams ended up as the data model behind a lot of later federated-social work, including activitypub and the mastodon protocol stack that came out of it. that lineage is real, even if the original opensocial gadget runtime never reached the audience the spec needed.

~ the decline ~

the cracks were that the member networks themselves were dying. orkut peaked in 2009 and lost its top spot in brazil to facebook by 2011. hi5 sold to tagged in december 2011 after years of declining engagement. myspace was acquired by news corp in 2005 then sold for $35 million in 2011 to specific media, a fall from a $580 million purchase price. friendster pivoted to a gaming site in 2011 and shut down for good in 2018. linkedin and salesforce kept their opensocial support around as enterprise-side features but never relied on it for their consumer pitches.

the facebook platform story was the real argument against opensocial. f8 had launched may '07 as a closed proprietary platform that any developer could build on top of, with one set of apis and one container that did not change beneath your feet. opensocial offered a standardized spec across many containers, but only if you wrote your application to the lowest common denominator across all of them. the developer choice was simple: build once on facebook and reach 50 million users, or write portably for opensocial and discover that every member container had its own quirks anyway.

~ what got buried with it ~

the opensocial foundation announced the w3c handoff december 16 2014. the standards work moved into the w3c social web activity, where it merged with activitypub-adjacent work and ostatus-adjacent work into what became activitystreams 2.0 and the federated social protocols that followed. the foundation itself dissolved. the opensocial.org website kept up archival documentation for a few more years then went quiet.

what got buried was the underlying premise that the social-web layer was something the industry would standardize around. by 2014 the answer was clear that it would not. facebook was the platform, twitter was the second platform, neither was interoperable with the other, and the standards bodies were where social-web ideas went when one company had already won the implementation race. the activitypub work later was a different attempt at the same goal, with a different generation of developers and a different deployment strategy. opensocial was the first big attempt from inside the industry consensus, and the failure mode it ran into was the failure mode every later attempt also ran into.

@}-->--

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

« previous  ·  random  ·  next »