Google Code

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

~ the obituary ~

i used google code around 2008 to 2010 for class projects and small personal scripts. honestly, ttnet broadband was working at home by then, university in ankara had me on a faster line during the day, and the question of where to put your source code in '08 had three or four answers depending on what your professor preferred. google code was one of them. it had subversion. it had a wiki you could edit without a build step. it had a downloads tab that gave you a direct link to a tarball, which was the thing you actually wanted half the time.

google launched it march 17, '05 as part of the company's open source push under chris dibona. the framing was always low-key: google had a lot of internal infrastructure for hosting source and running issue trackers, why not point a clean web interface at it and let the open source world use the same tools. sourceforge in '05 was already overloaded with ads and download mirrors that lied about which one was fastest. google code looked like the clean alternative for two or three years.

the issue tracker was actually decent. you could file a bug with labels, priority levels, the usual fields, and the search worked. for some reason this was hard to find in '05. the source browser supported syntax highlighting for the languages google cared about. the wiki used a textile-like syntax that nobody loved but everyone could read. the downloads quota was generous enough that a small project could host its release binaries for free. i remember pushing svn commits to a school project hosted there around '09, then walking to an internet cafe in ankara to verify the tarball downloaded for a friend on a slower connection.

github launched two months later in april '08, and the trajectory from there was set within four years. git rather than subversion, pull requests rather than patch files on a mailing list, a social layer that turned a repo into a thing you could star and follow. google code did not add a real social layer until it was too late. by '12 most new open source projects were going to github by default and google code was being used mostly by older projects that had not migrated yet.

~ the case file ~

~ the slide ~

the public reason google gave for the shutdown was that the site had become a spam target. anyway, by '13 a huge fraction of new project registrations were spam projects trying to game search rankings or distribute malware. the moderation cost climbed past whatever google had budgeted for what was always a side project for the open source group. in may '13 the ability to create new downloads was disabled because the downloads tab had become a malware-distribution surface. that was the real shutdown moment for anyone paying attention. the announcement two years later was just paperwork.

the deeper reason was that github had won the open source hosting market by then, and google had no incentive to spend more engineering on a service that no longer mapped to any product the company was monetizing. dibona's team built an export tool that would push your google code project to github or bitbucket with one click, which was honestly more graceful than most google shutdowns ever got.

~ what got buried with it ~

the issue trackers and wikis. that was the actual loss. the source code was easy to migrate because git supports svn import out of the box. the issue history was harder. a lot of older open source projects lost the historical context of why a bug was filed in '08 when google code went read-only because the migration tools handled commits well and metadata poorly. you can still find archived google code project pages on archive.org with the issue tracker partially preserved, but the search-by-label flow is gone.

the chromium project moved to its own infrastructure. android moved to gerrit. the smaller projects mostly went to github. a few died because nobody on the original team was paying attention enough to migrate by august 24, '15. for some reason the read-only window was only five months wide, which was tight if a project had been dormant for a year and nobody noticed the announcement.

~ the long tail ~

google still runs project hosting internally, and google open source still exists as an outreach program. but still, the public-facing project host idea ended with google code. github went to microsoft in '18 for seven and a half billion dollars, which was about thirty times what google would have ever spent to keep google code alive. the lesson, if you want one, is that hosting other people's source code is only a product if you have a business model attached to it. github has copilot now. google code never had anything attached.

@}-->--

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

« previous  ·  random  ·  next »