Ray Gans's Blog

Update on Mustang, Dolphin and the JDK Community

The last few months have been exciting for Java SE with the many JavaOne activities and presentations, our announcement to open source the JDK, and getting beta 2 out the door. Let me tell you, it's kept us all very busy here at Sun! Things aren't slowing down either, so I'd like to give you a brief update as to what's going on.

Mustang

Dolphin

Open Source

DLJ and jdk-distros

javadoc Translations

Survey

• Demographics: Info about the survey respondents
• Interests: General interests of the survey respondents
• Transparency: How to increase transparency
• Rewards and Recognition: How to improve rewards and recognition
• Community Building: How to build a stronger community
• Participation: How to increase community participation
• Java Use: How the survey respondents use Java SE

We will use this data to make improvements to the JDK Community site as we move forward.

Thanks to all who have helped us take Mustang to where it is today. Your code contributions, testing and bug reports have been outstanding and it has been rewarding for us here at Sun to work together with you. We're looking forward to improving our relationship with you as we get into Dolphin. If you're interested in joining or just want to see what we're up to, please take a look at the JDK Community of projects on java.net.

Comments

• Thanks for the status updates.

  Posted by: tjpalmer on June 28, 2006 at 07:54 AM

• I haven't been following the debian lists, but have they removed the packages from their apt repository? What's the issue there? And, are the .debs available on some other apt server?

  Posted by: richunger on June 28, 2006 at 11:41 AM

• richunger: the community, including Debian, had questions on the
  intent of the DLJ.
  Sun responded by clarifying our intent and that has helped significantly. So to answer your
  question the Debian packages are still available in the
  "non-free" archive and we have posted instructions on how to install
  them on Debian and
  Ubuntu.
  For followup questions on the DLJ or the jdk-distros project allow me
  to invite you to post in the jdk-distros forum.
  HTH, --Tom

  Posted by: tmarble on June 28, 2006 at 01:48 PM

• Mustang is looking great, can't wait to be able to use it in production.

  Posted by: firefight on June 28, 2006 at 03:43 PM

• Hi Ray,

  Thank you very much for publishing the results of the survey. tmarble was so kind to point me over here, and I am very impressed by your work on raising the bar for transparency here.

  I've looked briefly over the results, and I'd be interested in hearing what sort of ideas you think are worth taking away from it. My thoghts are below.

  I had some of my fears confirmed (Java(TM) still way to go to reach people outside the traditional places), but I think that's an area where the work done by us in the free runtime community around GNU Classpath (distributions, deployment, research, cross-language pollination, focus on both freedom & compatibility) can start paying off for Sun as well, once Sun becomes part of the free runtime world out there.

  It seems that the number of people willing to contribute per se is rather low (~5%). That is about what I'd expect. I think a good strategy would be for you to look at ways how to make it possible for Sun to look at working together with others out there in the same area, by enabling existing free runtime projects to share code with Sun's implementation.

  I was glad to see that a lot of people taking part in the survey understood that, too, and suggested mustering the courage to go the open source route, with a few people offerring some pretty specific advice on licensing, like keeping an eye on GPL-compatibility, which is a really big deal. Licensing seems to be a huge issue in the JDK community, and that should be high on the list of priorities.

  I was also glad to see a few suggestions to open up the TCK, and a few comments touching on the murky specification business and the JCP. I believe that's an area where Sun and independant implementors will have a lot of work to do to improve the level of quality of specifications and rebuild the JCP as a better place, that actually represents the interest of most Java developers, instead of what it is now. Current survey on java.net seems to indicate there is a long way to go there. Webmink cited the CORBA downfall document on his blog somewhere, and I think that one has a lot of lessons to learn for Sun in the internal governance debate. I've promised Simon to say more about that on my blog sometime, so I'll leave it at that here.

  I was also happy to see 46% of participants say that "Closer ties with open source development efforts" are "very important", which is in my experience really, really important. GNU Classpath really started to take off once we began to weave a tighter net between all the different efforts occupying different niches, from distributions, to VMs, to runtimes, to utility and library projects we could use, or that could use a free software implementation of Java. Come and join us next FOSDEM.

  Some of the other results are similar to the lessons I learned maintaining Kaffe: a broken build won't help people submit patches, even if it works for you/your team, for example. I was surprised at how long it apparently took to have people up and running with their own builds once you started publishing the sources to mustang, for example.

  Finally, I'd like to say that the JavaDoc translations idea is a very good one. In particular, it's great that even the holy cow of API specifications is no longer safe from being improved into something more useful. If you have some time, check out the GNU Classpath JavaDoc for some ideas on how the specification could be improved, like http://developer.classpath.org/doc/java/awt/geom/CubicCurve2D.html for a colourful example, generated from GNU Classpath's source code by gjdoc. I hope Sun will make it possible to get the sort of improvements into, say, an open source spec for Dolphin, which will require licensing the spec's javadocs under a sane license, of course, to be able to share them between class library implementations.

  Posted by: robilad on June 28, 2006 at 04:15 PM

• 
  Thanks Dalibor for your comments and suggestions.

  The main purpose of the survey was to help us learn if Sun was far off the mark in what we've been doing with the JDK Community. I'd say the answer (for most people who responded) is that we are doing OK. This is good to know, so we can concentrate on its secondary purpose, which was to identify pain points and needed improvements -- many of which are obvious from the results and comments.

  For example, better transparency demands we make more information available about new developments and ideas through blogs and articles. I've been really pleased by the blogging that Java SE engineering has done this past year and expect it to continue, but the survey also shows that collecting this information together by development team (via webpage/newsletter) would be highly desireable. Providing better status on bug reports and contributed fixes are also things we have to improve. We're looking into this.

  The same goes for the other focus areas (Building Community, Rewards & Recognition, and Participation). I'm not committing that Sun will address all the issues raised, but we are trying to understand them and use what we learn to better prioritize our next steps.

  Another area we'll be looking at soon is how to better engage with the community that lives and works in an open development environment so as to listen and learn from their experiences. The survey is one step in that direction, and I expect we'll be doing much more in the coming months.

  Posted by: ray_gans on June 29, 2006 at 05:00 PM

• Thank you for taking the time to reply, Ray, and offering your insights. I'm looking forward to continue the discussion in the coming months. If you can spare the time, please keep blogging, I think quote a few people would find information showing thought processes inside Sun on the way to openness interesting.

  cheers,
  dalibor topic

  Posted by: robilad on June 29, 2006 at 10:31 PM

• Thanks for the update Ray!

  
  I was sad to see that the "Java VM always running as a service" feature was dropped from Mustang :-( PLEASE make sure it becomes part of Dolphin! It is THE powerful way to reduce start-up time and make Java apps and applets feel like native. AFAIK, it was to be included in Mustang, but then someone wrote a lame excuse why it was removed. That particular entry generated a lot of comments questioning the validity of the excuse.

  
  We need an always running VM as a service please. Most other issues with Java are dwarfed by this. (If you somehow still included it in Mustang, that would be great, but even that would be late!)

  
  Thanks!

  Posted by: atus on June 30, 2006 at 06:12 AM

• I really hope that the JAva running as a service all the time is an option and not enforced by default. It would truly be annoying to have it eat a constant 150 mb of memory all the time!

  Posted by: suryad on June 30, 2006 at 09:01 AM

• a) a service is always optional
  b) Java itself doesn't eat 150 mb of memory
  c) finishing and integrating isolation API (a prerequisite for Java-as-a-service) is extremely desirable, not only for desktop Java apps, but for mobile, and especially enterprise Java (it would be extremely helpfull for Java hosting)

  Posted by: selendic1 on June 30, 2006 at 09:38 AM

• Ray, please see to it that access to the sources becomes more useful than having to download a huge bundle every week. I asked about a CVS/whatever access months ago and was assured this would happen eventually: http://forums.java.net/jive/thread.jspa?threadID=2547 . Matthias
  

  Posted by: mernst on July 03, 2006 at 10:19 AM

• Ray;
  I am happy to see mustang mature but I am very sad for two reason.

  First, where is RAW Socket support ? This is still a key bug since years ! Java is supposed to be very network friendly but there is still NO Java based solution to send IP packet that are not TCP or UDP related (taking aside the new "ping like" facility) !!! This is really not understandable for me :(

  Lots of key projets are depending on this to bring new networks behaviours and implements advanced protocols and services.

  Second, I made two contributions to Mustang and I got still no news on the status of integration. This is realy sad for me as I spent quite a time to polish this and make it sure it would fit in Java SE RI. If you plan to make the community think Java SE is a project that worth spending time on it, you realy have to rework the contribution tracking system. We do need something that will show the validation process and the current status for a given contribution ... "good old bugparade" is no more suited to handle modern issue tracking. Ask the GlassFish team, they have shown good community interraction on the V1 and this was a win-win situation IMHO !

  Rgs,
  JB

  Posted by: bjb on July 04, 2006 at 08:16 AM

This work is licensed under a Creative Commons License.