Thursday, June 18, 2009

Kill Project Jigsaw Now. Please!

We need to relieve poor Alex Buckley from his suffering. What he is complaining about here is due to the fact that he could not answer the very first (and most important) question put to him by the OSGi expert group: "Why didn't you start with OSGi?"

Furthermore, during that morning's debate, the OSGi members managed to completely invalidate every technical justification for why it's necessary for Project Jigsaw to reinvent OSGi (it clearly isn't).

To be fair, Alex has been put in a very bad position as JSR 294 spec lead. It's widely accepted that modular programming could be improved through extensions to the Java language, and the work in this direction to date on JSR 294 has been well received (with the exception that it has not defined a mapping to OSGi of course). But the problem for JSR 294 going to continue to be that there is no technical reason for mapping to Project Jigsaw instead of OSGi.

It would be so much easier on Alex -- and on all the rest of us interested in Java modularity -- if we could just kill Project Jigsaw. And the sooner the better.

It would also save James Gosling from making further embarassing statements about OSGi. (OSGi is not fat, is not big or bloated, and were Project Jigsaw to live it would have to duplicate most of what's in OSGi anyway.)

There are other good reasons for killing Project Jigsaw. But the main reason is that the OSGi specification already defines a perfectly good, mature, tested, and widely adopted modularity standard. It's been around for more than 10 years, has a JSR and an open process through which the leading Java vendors participate actively (unlike Project Jigsaw), and has gone through several major iterations (R4.2 is very nearly completed as well).

OSGi already has been incorporated successfully into thousands of projects and products, and is endorsed by nearly every Java vendor (including Sun, or should I say at least the part of Sun working on Glassfish).

The industry does not need another Java modularity standard. That's like having two competing alphabets from which to compose words. It just doesn't make sense. It is counterproductive, especially when the IT world needs to focus on reducing cost and improving flexibility. No wonder Alex is pulling his hair out.

The Jigsaw guys had their 15 minutes of fame at Java One. For Alex's sake, and the sake of everyone else involved in Java modularity, let's kill it. Now. Please.


  1. I loved the characterization of the meeting by Alex as "a verbal mugging". I always find it fascinating when people bitching about behavior instead of arguments when they find they have no arguments of their own.

  2. Yes, I totally agree.

    But that's why someone should take pity on him and eliminate the problem.... ;-)

  3. My understanding is Sun wanted to move fast as they needed Jigsaw to allow modularization of the JVM (desperately needed for Java FX). They wanted to move independently as moving within standards bodies and committees is inherently slow. So they did their own thing. Fine by me, few actively use OSGi yet anyway.

  4. Funny how you have to be anonymous...

    "...few actively use OSGi..."

    What planet are you on? OSGi is used in more than a thousand technologies and products, and every Java vendor (including Sun) has adopted it.


  5. ps to the comment about moving slowly within standards bodies, OSGi is already done, there was no need to move slowly. Developing a new standard, whether in committee or not, is by definition slower.

  6. discusson on jigsaw/osgi

  7. I have not yet listened to the podcast which is the basis of the discussion Phil mentions (I will shortly) but it sounds like the Jigsaw folks must be at least consistent in the misinformation they are giving out, since it seems they must also have said similar things to James Gosling. See Chris's excellent rebuttal: