
JXTA
» see jxta.org as it lived, on the wayback machine
~ the eulogy ~
In 2001, on a modem, peer-to-peer meant one thing and it was not architecture. It meant a file you wanted sitting on a stranger's machine, and an evening spent watching a progress bar while nothing else on the connection worked. JXTA was the other kind of peer-to-peer, the kind written for people building software rather than people downloading music, and from where i sat it was completely invisible.
Sun Microsystems released it on 25 april 2001, with Bill Joy and John Gage presenting. The name is short for juxtapose. The pitch was that peer-to-peer should not be a category of application at all but a layer that anything could sit on top of. The wire format was XML and the addressing named peers instead of machines, so a device could keep one identity while it moved between networks.
It was ambitious the way 2001 was ambitious. The specification did not care what language you wrote in or what carried the bytes, and there was a C port next to a cut-down version for handsets. Academic papers about it kept coming for a decade. What never arrived, at any point in those nine years, was a piece of software normal people used that happened to be built on it.
Oracle finished buying Sun in january 2010 and announced its withdrawal from the JXTA projects that november. Volunteers put out version 2.7 in march 2011, into a room that had already emptied. There was talk of moving the code to Apache and nobody at Oracle answered. jxta.org does not resolve today, and the project's last home went dark along with the rest of that hosting.
~ the dossier ~
| Born | 25 april 2001, source and tools released by Sun |
|---|---|
| Peak | a Java reference implementation with a C port and a cut-down version for handsets |
| Killed | november 2010, when Oracle announced its withdrawal |
| Lifespan | nine years |
| Made by | Sun Microsystems, presented by Bill Joy and John Gage |
| Killed by | the Oracle acquisition, and a web that answered the same question with servers |
~ what it promised ~
The idea underneath JXTA is a good one and it keeps coming back in other clothes. Give every peer a stable identity that survives a change of network, and let a message find its own way there through whatever path exists rather than through a fixed server in the middle. That is close to what a modern peer-to-peer stack does when it works.
The difference is in which end you start from. The stacks that survived were written for one specific job somebody wanted done, and the protocol got generalized afterward by people who needed it to do a second job. JXTA was built as the general platform first and then waited for the jobs to come to it.
~ why nothing shipped on it ~
The timing was wrong in a way no amount of engineering would have fixed. In 2001 peer-to-peer was the most interesting subject in computing, because Napster was being sued to death in public and every large vendor wanted a position on the question. By 2004 the answer to nearly every distributed problem was a datacenter with an HTTP endpoint in front of it, and it stayed that way for fifteen years. A protocol suite for making machines talk without a server was solving a problem the industry had decided to solve with servers.
The other complaint was weight. The reference implementation asked a lot of a machine, and the abstractions cost more than they returned for anybody whose actual requirement was two computers talking to each other. Developers who tried it tended to come away with a working demo and no reason to keep going.
~ the long tail ~
Sun kept the source open the whole way through, so the code is still out there for anyone who wants a museum piece. Grid computing research kept citing it into the 2010s, mostly because the papers were already written and a citation is cheaper than a rewrite. That is the afterlife of a platform nobody adopted. It survives in the literature about itself.
What JXTA left is a lesson about the shape of a launch. It arrived as a complete architecture with a specification and an implementation, and without one application anyone needed. Nearly everything that won in the same territory arrived the other way around, as a single working program whose protocol got documented later by people who had already started using it.
~ leave a rose ~
an anonymous tribute. one rose per visitor per day.
@}-->-- ... roses left here
~ leave a wreath ~
a rose fades by morning. a wreath stays on this grave for good, with your name on it.
from $1, paid once through stripe. it keeps the graveyard running: the hosting and the nightly digging. adopt it and the headstone reads "maintained in memory by" you. the groundskeeper plants new wreaths at the next tending, usually within a day.
» one email a week: new graves and the anniversaries of the dead. subscribe
other protocols: dns0.eu, Dukgo, Mozilla IRC, SPDY
also died in 2010-2014: Itsmy, Legends of Zork, Infantry Online
~ leave a note ~
visitors before you have left these graveside notes. anonymous welcome.