Tuesday, 18 November 2008

Eclipse Summit Europe 2008: The Members Meeting

After a good night's sleep, Monday was the start of beautiful sunny day. We did some sightseeing in town to begin the day.

In the afternoon, I attended the members meeting. There are now roughly 190 members, 22 of which are strategic members, and there are 932 committers from 75 organizations. These days we get roughly 1,000,000 package download requests per month. We are expecting about 400 ESE attendees and there are roughly 35 sponsors.

Mike talked about the initiative to create industry working groups as a way to drive the growth of the foundation and to diversify its membership. Eclipse has good infrastructure and processes, including a sound intellectual property process, to help facilitate collaboration. The working groups can leverage existing projects, feed back their requirements to those projects, and even initiate projects to address any gaps. Our first example is the Mobile Industry Working Group led by Motorola and Nokia which will focus on creating a Model Application Development Kit.

There were questions about licensing issues for these working groups, i.e, will they be stuck with EPL? Mike pointed out that 3 projects are already dual licensed, e.g., EclipseLink is also licensed under BSD so that it be consumed by GPL and be part of the JPA reference implementation. He also indicated that there are efforts underway at the board to make it possible for something like CDT's Wascana distro to be hosted at Eclipse.

Karsten Schmidt talked about the SAP Eclipse story.

SAP decided to support Java in 2000. They chose Borland's JBuilder as their IDE but they needed better access to source than was provided by JBuilder in order to build their own specific extensions, so they evaluated Eclipse and NetBeans. They also needed a model abstraction layer. By 2003 they had a version of NewWeaver based on J2EE 1.3 with Selena as their modeling layer. At that time, Borland acquired TogetherSoft and Modeling Infrastructure (MOIN) based on MOF 1.4 was started as a research project. They decided not to use EMF because they needed something that worked on the server side; how odd given that EMF has always worked there. By 2006, they became the first major vendor certified for JEE 5. By 2007 they first signed the Eclipse committer Agreement and supported a committer on WTP. They also donated their memory analyzer project to Eclipse and they now have a BPMN modeler based on MOIN.

Some challenges they faced included their conservative policies that kept them from participating; without participation they could not have real influence. Things like the EMF/MOIN split are an unfortunate consequence of that lack of direct involvement. Karsten expressed regret that they were not able to get directly involved with EMF many years ago; I was certainly very hopeful after last year's "Big Models" meeting that we'd be able to work more closely together. Unfortunately decisions made at the executive levels don't always match the wishes or expectations of those of us in the trenches. Karsten is still hopeful that perhaps it's possible to better align EMF and MOIN in the future. They also expect to fully align their current component structure onto a uniform OSGi base, and they plan to focus on the Equinox implementation for that, because the benchmarks for Equinox generally give it the lead and therefore make it the best choice. The fact that it's hosted at Eclipse makes it easier for them to participate in its evolution. It's a brave and bold new future for SAP and Eclipse.

Jeff McAffer presented an overview of the runtime project's vision of the future.

The trend these days, as Karsen just demonstrated, is that people are more and more willing to consume from open source. One problem this surfaces is how to combine those components to build solutions. Many things are possible, including the ability to install a thin Equinox WAR into which you embed your OSGi bundles. The Utopian vision is to be able to use the same component model and deploy it anywhere. We already have all major JEE vendors building on OSGi, although they don't surface that as a component model for their users. A consistent, flexible, configurable component model that allows you to integrate a diversity of components is key.

Next Jochen Krause presented on the motivation behind the e4 effort and the progress it's made.

It needs to be easier to build plugins including doing so in other languages like Javascript. Richer more flexible support for styling is an important goal, along with a service-based application model able to target the web. He wanted to reassure the community that we will remain committed to the 3.x stream for five years or more, and that we plan that 3.x plugins will continue to work in a compatibility mode in e4. We will strive to provide more uniform APIs that avoid singletons and are more finely structured as small components. Even the resource model needs to be more flexible and we need to exploit patterns like Dependency Injection. We expect that the entire workbench will be formally modeled. Overall, the goal is to create a simpler and more powerful application model. An interesting approach that has been prototyped is cross compiling SWT to Flex. He also showed Boris' sample e4 application running normally and running as a brower application; pretty impressive! Of course of invited folks to get involved!

Chris Brockman presented on some of the problems that need to be solved in the business world.

The integration of disparate sources of information across the enterprise is key issue to many. The same information can be interpreted in different ways and often separate sources of information need to be related as if they originated from a uniform semantic source. The SMILA approach is focused on this. The idea is to create a shared architecture standard. Just a single company like Volkswagen might have 500 applications running just in Germany, so even just maintaining connectors for these applications is a significant cost. It's easier for application based on Eclipse because providers can be more easily encouraged to provide integration directly. Important players like SAP have gotten involved. He also drew attention to Eclipse's IP process, which, although can seem onerous, is something that's key to using open source in a way that's enterprise friendly. Ultimately it has extreme value for corporate consumers.

To help bring in the schedule, Ian Skerret gave a very rapid overview of the Mobile Working Group Mike talked about earlier.

It's a great example of the foundation's successful new approach for attracting new types of members.

Martin Oberhuber presented an overview of the Architecture Council, what it is, and what it's doing.

There is a representative for each PMC, a representative for each strategic member, as well as members appointed based on merit. The AC council facilitates mentorship and all new projects must have at least two mentors from the AC. It fields community requests via bugzilla; did you know you could do that? It interacts directly with the EMO. And it acts as a channel for raising high level issues to the board. An example of how an issue was addressed was the change to make the committers mailing list be moderated so it can remain open without becoming a frustrating source of SPAM. The AC can help offload work for the foundation staff by fielding questions directly; we have a lot of experience to share. In general, its overview of the community helps it act as a filter when it comes to issues reaching the EMO and the board. Martin has done an outstanding job making the AC active and effective.

Our beloved Bjorn, acting as Rich, talked very briefly about the activities of the Planning Council,
which has recently set itself up as the body to drive the release train for the Galileo release.

He talked about the impact of the tragedy of the commons, i.e., how everyone might want something but no one is wiling to help make it come about, with the hope others will do it for them. See how he's going to crush those who don't pony up? They don't call him knuckles for nothing. We fully expect that Galileo will release on time and on schedule just as every previous release.

Maxim Groouchevoi introduced instinctools, a new Eclipse member based in Belarus.

He talked about the DITA technology developed by his company. The idea is to provide single-source support for the documentation process and to support the full documentation life cycle.

Finally Ralph talked about the growth of Eclipse in Europe. There he is explaining how he's carving up the European pie.

All this talking and listening is hungry work so a group of us stormed a local restaurant in search of a cozy table for 22.

It was a long hard day, but even the walk back was pretty.

I was looking forward to the symposiums!

No comments: