java.net

» see java.net as it lived, on the wayback machine

~ the obit ~

Java was the language my department taught in, and for a long stretch of the 2000s a Java question typed into a search box would put a java.net link in front of you. A project page, or a mailing list thread from three years earlier where somebody had the same stack trace. I read that site constantly without ever hosting anything on it, which was the normal relationship. Most of its traffic came from people who arrived at one page, took what they needed, and left.

Sun announced it at JavaOne on 11 June 2003. The pitch was a community website with a source code repository attached, so that Java work could happen where anyone could watch it happen. The same conference launched Javapedia, an encyclopedia of the Java platform that ran on TWiki and let only registered users edit. That one went nowhere. The forge underneath it did not.

What made java.net matter was not the code hosting, which was ordinary. It was that the Java Community Process ended up keeping its public record there. If a JSR had a project site, a reference implementation, a bug list, an argument about how a specification should read, all of it lived at java.net. A standards body had put its papers on one vendor's hosting.

Oracle bought Sun and in January 2010 said it would migrate java.net onto the codebase behind Project Kenai, its other forge. Oracle had spent part of that year telling people Kenai was closing and then walking it back. Anyone paying attention had already learned something about how long these addresses were guaranteed for.

~ the rap sheet ~

~ what it hosted ~

The visible half was project pages and downloads. The half that turned out to be irreplaceable was everything around them: years of mailing list traffic, issue trackers with the reasoning written into the comments, expert group discussion that explained why a method signature ended up the way it did. Specifications tell you what was decided. Those archives told you why.

~ the actual end ~

In June 2016 Oracle said java.net and kenai.com would both go dark on 28 April 2017. The reasoning given was that developers had already consolidated onto a small number of popular platforms, GitHub above all, and that Oracle could add more value by investing there instead. As a read of where open source had gone by 2016, that was correct.

As a shutdown it was thin. There was no exporter and no offer to move anybody's repository for them. Reza Rahman of the Java EE Guardians pointed out that Oracle was asking every project owner to do the move themselves, on their own time, for infrastructure the Java EE specifications depended on. Some projects moved cleanly. Others moved the code and abandoned the history, because the code was the part with a button for it.

~ what got buried with it ~

Thirteen years of links stopped resolving on the same day. Every answer on Stack Overflow that cited a java.net thread and every blog post pointing at a JSR project page now ends at nothing. The software survived because software gets mirrored. The conversation around the software did not, because nobody mirrors a mailing list they do not own.

The lesson is not that Oracle was cheap about it. It is that a process built to be open kept its only public record on a company's servers for thirteen years, and when the company was finished the record went with it. GitHub is somebody's server too.

@}-->--

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

« previous  ·  random  ·  next »