As some of you might now I have started using Twitter for some time now and writing tweets since a view weeks now (http://twitter.com/goertsch). As most of the people I know I am not using the twitter website directly, further more I am using TweetDeck at the moment because it is the best application I found until now. On the other hand I gave the new friendfeed interface a try for the last two days. The new beta of friendfeed is very interesting because all components are updating in real time. This gave you the impression like you have a chat window open where all the different information is pooping in from the different services.

I really like the live messages and are following them them very often, especial at conferences e.g. the EclipseCon I visited a view days ago the medium is very useful. On the other hand I don’t have the time to follow the live messages all the day and special because I am living in Europe I miss moste of the messages in real time. This is really an issue because all clients I know are focosing on the live updates and fancy dynamic content that is updated nearly the same second the person is hitting the enter button on his keyboard and are not taking care about people who want to be able to read the content in a more ‘offline’ way.

I am looking for a interface that is more like gmail. It should of course support keyboard navigation and give me feedback about information I have already read. I have just downloaded Adobe Air (TweetDec is developed on it) and will try to prototype an application.

KeyNote - Building Applications for the Cloud with Amazon:

Was an interesting Keynote. Most of the stuff was not really new to me but at the numbers are always impressive. SmugMug presented their Solution which is really impressive. They are doing all their processing in the cloud by amazon and are very happy with that. The interesting part is that they are just acquiring resources (servers) as they need it. He spoke a little bit about their solution and how they have designed their Infrastructure. An server checks periodically how many requests are in the queue and what kind of they are. If they decide that they need more resources they are just starting an or more server instances at amazon EC2. The server is an plain instance of linux an just has some configurations that after starting he calls home what he has to do. If there is still work to do the server provides him with information where to find the software and what to execute. After the work is done he makes the same call. If there is more work to do the server receives the new work item and new software if needed. If not he shuts itself down. Nice concept and very easy to maintain.

After the presentation of the SmugMug CEO amazon announced their new AWS Toolkit based on Eclipse which is really a grate tool for managing servers in the cloud.

Presentation

Building Communities: The Rise and Fall and Rise of the CDT Project

Doug Schaefer gave a grate talk about building a community and how to help them to stay on live. One issues he had noticed is that you can’t just go and say commit. This would not work out because there are to many people and especial to many bosses involved, furthermore if you want the things to happen than you have to contribute and not rely on others.
Prepare to be frustrated because it is very frustrating to build a community, for every vendor or person that contributes there are four just taking the stuff and are happy with it.
It is important that as an committer in the professional area you have to wear two hats. You have to make your vendor happy and also work for the community. Asking people is not enough, you have to create the need and meet face to face with the people to have the chance to build an hold a community. Communicate, communicate, communicate and be honest to the people and if you have a hard time to get people convinced maybe they are just not interested in your idea and you have to rethink it.

Presentation

Building LinkedIn’s Next Generation Architecture with OSGI

This presentation first of all showed me that applications are growing and some of the most used sites are just a mess. LinkedIn is based on some war applications but most of the logic is in one file located. Because this became a maintenance problem they decided to come up with an OSGi based infrastructure where they extracted the services and tomcat was just one container within the OSGi container. It became a problem that they had so many dependencies between the different bundles now that they had to start refactoring. The nice and bad thing with bundles is that they make the dependencies visible because you have to declare them.

Presentation

Eclipse @ eBay

Michael Galpin did a great job explaining the history behind the eBay Website. They build it about 4 times, where the first one was written in Perl and did not have a database. After that they tried an c/c++ architecture. 2001 they started to replace it with Java. They used an XML/XSLT Layer for the frontend, but that did not work out for them therefor they decided to switch all to Java. Of course Java is not the Silver Bullet, but they tried to bring the all information on one level and use code generators to represent the information.
What was really interesting to see was the issues companies with an service oriented architecture are facing. Because of the high amount of different projects/services they needed to write software that is setting up the desktop of a developer in a way that he can work with, otherwise it would take him days to find the correct configuration and plugins. Because of that they stick to eclipse as their platform and wrote plugins to set up the environment. They have also written other plugins for eclipse one of them is a java script editor which mabye announced at the javaone ;-).

Presentation

Working in the Cloud: Development, Deployment and Project Management for Force.com

Was more an advertisement of force.com than technical information.

Presentation

Symmetric Service Oriented Programming

Was an nice presentation about OSGi and dependency injection and how to combine them.

Presentation

I was really impressed what the Red Hat guys have done to make C/C++ programming easier. A few years ago I developed SW in C++ for Linux and it was a real pain to memorize all the command line option for all these (useful) development tools. The Linux Distros/Tools integrate some of these tools into eclipse. E.g. the integration of valgrind really succeeded. And if you ever tried to write a rpm script without syntax highlighting and code completion you will be very impressed by this project. It’s a little bit ironic that this useful C/C++ tools are implemented in Java

The Glassfish developer make a big effort on the modularity of the server. And if I can trust their talk, they made a great job. You have only to include the modules you realy need. Glassfish v3 uses OSGi for module management and all Glassfish modules are OSGi modules. But it doesn’t directly depend on OSGi. The service mapping is done via HK2, an other open source project.
For the next project I need an application server I will have a closer look to Glassfish.

It’s a very interesting possibility to use the same code base for RCP and Web applications.  But this will not replace RCP applications, and single sourcing is of course not for free. You have to divide your code in common part, a RCP part and a Web part. This can be done via bundle fragments (and the delegate pattern). This reminds me a little bit to

#if defined (WIN32)
 ...
#elif defined (LINUX)
...
#endif

blocks in C++.

The software-update-button for example doesn’t make sens in an Web application. As the talk points out, there is also no graphic context within a Web application.  So you can’t draw arbitrary shapes, as it’s possible in an RCP application. There is also no possibility to catch every mouse-move event, which will provoke heavy network traffic to the server.

The conclusion for me is, that it is a very interesting possibility to reuse code, but not the silver bullet, as some people think (or hope).

Designing an Android Domain Specific Modeling Language using EMF, GEF and GMF:

I was really impressed how easy it is to develop an application for Android with the modeling tool they build it is really strait forward to build the interaction within an diagram. I really would like to compare the platform of the iPhone and the android one. Let’s see which of them will hype more in the future. I can see potential in both of them.
On the other hand i found some things concerning EMF we where doing wrong in our own implementation. Will have to take a closer look at that as soon as i am back in Vienna. For further information i would suggest to take a look at:

Presentation
developer.androide.com

Executing BPMN

Was an nice talk about BPM and all the acronyms that twitter around in this area. At least the presenter was realistic in some way as he mentioned “it will never happen that a business analyst can design an WF and deploy it without any developer”. I totally agree with that and have to admit that i we should have taken this more into account while designing the CEP system at Senactive. Anyway i am still confused by all this different standards and quasi standards (BEPEL, BPMN, jBPM, ….).

Presentation

e4 Project in Review

In my opinion this was one of the most interesting sessions i visited at the conference. It showed at least some of the details in action that e4 will change in eclipse. In particular they showed the new Resource framework and XWT with the Model behind the Workbench. They are also going in the right direction by removing the whole Singleton chaos that exists in the current WB to provide the possibility to use the RCP in an multiuser environment. This in particular is needed to give better support for the RAP project which by the way will also be an essential part for the e4. e4 is still in the incubator but will release an first version in July this year and hopefully ship in Summer 2010. As there are several advantages with e4 we will still have to wait for about 2 years to get a stable version that we can use for RCP development.

Presentation

Make Money Fast! Market Your Killer App with the Eclipse Pulsar Initiative (formerly Eclipse Mobile Industry Working Group)

This was an interested discussion about the Pulsar initiative that was announced some time ago. Some mobile companies including Motorola, Sony Ericsson and Nokia are trying to build an environment for mobile developers based on the Eclipse Framework to support an easier way of development for their devices. For me it sounds like they want to form abase against the two other big new players in the market, Apple with their iPhone and Google with Android. The discussion itself was a little bit boring and i am not sure how this will work out in the future.

Presentation
Announcement

I haven’t understand people twittering around until now. Arrived at EclipseCon 2009 I looked around and found everybody twittering. I had an account there for some time now and because I had time and wanted to try what it means to twitter i started. I’m doing this now for two days and I really have to say i see the advantages of the system. The website itself is not really very useful but if you use tools e.g. TweetDeck that supports you browsing through the messages you really get Information fast and in some kind of sorted way. For me the interesting thing is that you don’t have to write whole stories like in a blog.

If you want to follow me on twitter just go to: http://twitter.com/goertsch

This talk give a short overview how eclipse evolves from an IDE over a RCP to the server side. As a conclusion you can say, that the “cloud” and “p2″ is everywhere. On top of p2 you can (must) build your own SW deploy system.

Not the sexiest part of the eclipse ecosystem, but everyone who is doing something useful with the eclipse RCP has to deal with the eclipse resource framework. The resources in the workspace are represented as tree. Eclipse makes a big effort to make the cost (eg. for rebuilding the source code) proportional to the changes  and not proportional to the size of the workspace. This is done by tracking the differences.

What I missed (or ignored) so far was the concurrency issue. The resource tree only exists once, so you have to deal with concurrent read and write access. For read access you don’t have to aquire a lock. Write access uses the “copy-on-wirte” pattern. For a more advanced locking you have to use the ILock interface from the eclipse jobs framework.

As the title let you mention the tutorial was about bringing an Eclipse RCP into the web. The idea of single source for desktop and web is not new, but in the last time more and more projects start providing a framework that may really work in the field.

Rap is providing a bridge of the SWT shell to the web. All eclipse.swt controls (at least they told me) have been ported to the web in the rap packages. Others e.g. nebular are still missing. Supporting your control to be able to render in the rap framework, which by the way can be done by everybody, is hard work to do because you have to implement the web representation and the javascript code for the communication. In RAP the client (browser) does not know anything about the state and the processing logic, this code is just running on the server.

As Eclipse in its current state is just a single user application there are several issues. Until now developers don’t have to  to care about the context of the user because there is always just one instance running with one user running. Eclipse used Singletons heavily within the RCP and so did developers. Of course because RAP is just starting one instance of the Eclipse RCP this have several disadvantatges and will in several cases lead to illegal states of the application (e.g Shopping Basket where several users would just access one instance). This status will change with the e4 which is in the pipe but not shipped until now.

Most of the code can be used in both worlds the RCP and the web but some is only needed or hast to be changed for one of them. Because of that they used the OSGi Fragments to support the functionality of replacing information depending on the client they are running on.

For the web they did a grate job to allow styles and layouts that the application does look more like an web application then just an RCP within the browser. See the two screenshost below to find out what is possible with thees techniques.

Future: For the future they try to define some kind of generic protocol for the client server communication and to think about using gwt to generate code for their client controls.

Conclusion: RAP did a grate job in my opinion nevertheless I have to say that there are still some weaknesses. For me the project makes sense for smaller RCP applications but not for complex ones. Sadly moste of the applications we are speaking about are the complex ones.

Presentation