SDA: Peter, how do you feel about being nominated the "OSGi Evangelist"? What is the main goal you want to achieve?
Peter Kriens: It is exciting to be able to promote the technology I have been working on since 1998. Working with some of the best Java experts in the world and writing the specifications has been a fantastic experience, but now it is time to turn this opus into the success it deserves.
We have lots of plans for the coming year, and some parts are already in progress. First, the OSGi blog is going to get a new article written about OSGi technology twice a week. This is in addition to my private blog at www.aQute.biz. In the coming months, I will use the blog to show how you can make your own OSGi server run on a PC to perform all kinds of useful tasks while taking advantage of the many R4 features.
We will also start to host a bundle repository on bundles.osgi.org. I have been working with Olivier Gruber, Jeff McAffer, and Richard S. Hall (who pioneered the Oscar Bundle Repository) to set up a federated database of bundles. Any serious group that has bundles, both commercial and open source, will be able to participate in this scheme. This will hopefully create a market place for bundles.
For the embedded market, we will provide a demonstration platform based on the NSLU2 from Linksys. This is a cheap device (US) that can run OSGi. I hope to develop several bundles for this constellation, showing that OSGi is not only viable on small hardware, but also that it provides a clean management model.
At the recent ApacheCon, I discussed with several key OSGi people how to handle the lack of good OSGi programmer documentation. One of my highest priorities now is to set up a table of content and whip the writers in submissions. We hope to have such book out before summer.
So, it will be a busy year, especially because we are also still working on the Mobile and Vehicle OSGi Service Platforms, and I have not been relieved of those duties.
SDA: The OSGi R4 Core Spec is public since the OSGi Congress in October 2005. What are the major improvements over R3 for you?
PK: Declarative services, Event Admin and the new module features.
Declarative services really simplify the programming of OSGi bundles.
It is scary that outsiders (Humbert Cervantes and Richard Hall) had to come up with Service Binder to make us see this idea. We could have been spared a lot of grief if we had this in R3; I always hoped that I was immune to this kind of blindness, but alas.
Event Admin provides a central publish/subscribe event model. We had been talking about this for a long time but this time the Mobile Expert Group pushed us over the brink. Event Admin reduces the need for many different listeners by providing a central point for all event management. It significantly simplifies many programs.
The biggest overhaul in R4 has been the module layer. We now support a module layer that supports multiple versions of the same packages. R3 class sharing was designed to share specifications, not implementations. In R4, we fully support implementation sharing as well.
SDA: Which publicly available (open source?) OSGi R4 frameworks exist? What is the status of the standard services implementation?
PK: Open source implementations are (by my knowledge): Apache Felix (formerly known as Oscar), Eclipse, and Knopflerfish. All three are currently working hard to provide the standard services, but I reckon this will take some time before the more specialised services are available. Then again, Knopflerfish has fully implemented R3 and are working hard to make them R4 compliant. These bundles must work on the other frameworks as well. The biggest missing part is declarative services. Only Eclipse has an implementation so far. Unfortunately, declarative services require a backdoor into the framework, making it not straightforward to use on other frameworks.
Companies like IBM, Espial, Atinav, ProSyst, Gatespace Telematics, and others provide commercial implementations of the services.
SDA: What can mobile developers expect from Mobile Experts Group Framework (also known as the OSGi Mobile Framework) and will major handset manufacturers support it?
PK: Mobile developers can finally escape the sandbox that MIDP is holding them hostage in. The OSGi Framework is a much more mature and system-oriented environment, not restricted by the game focus of MIDP. OSGi-enabled phones can be used for enterprise applications and system services. The service-oriented architecture of the OSGi Framework enables applications to share, collaborate, and adapt to their environment. Key is also that finally Java can play a role on the system level; the Mobile Platform specification clearly enables writing system components in Java. For example, screen managers, OMA DM node implementations, deployment resource type processors, and much more.
I am promised a Nokia phone with a preliminary version of the Mobile Service Platform. Of course, the key issue is how the operators will react. To be successful, we need an operator to see the business opportunities that OSGi-enabled phones provide. Once one operator can produce success stories, I have no doubt both manufacturers and operators will scramble to get a seat on this new market of mobile enterprise software. I see very keen interest in some of the large players in the enterprise market, and it would not surprise me if Nokia makes a lot of money in this market. Though the numbers will be much lower than MIDP sales, the margins in the enterprise markets are many times higher than on Midlets.
SDA: If you could turn back time to 1999 when the OSGi Alliance was founded, what error would you NOT commit again?
PK: Thinking we would be an overnight success. I think we were a bit too ambitious in our quest to change the software world. OSGi programming had a lot of benefits, I still think the dynamics are wonderful, but the programming model was not easy for newcomers.
Declarative services has now changed that, but I think we should have started with a model that did not require another way of thinking. But, I naively believed that we could change the world when IBM, Sun, Ericsson, Motorola, Oracle, Nortel, and other fortune 100 companies put their weight behind us. I now know better.
SDA: What can we expect from R5? Is the work on R5 already underway?
PK: The board has decided not to start any major work on R5 and instead focus on getting widespread adoption of R4. However, we obviously have a list of items we would like to work on. Still, 2006 should be the year of adoption. Looking at what is happening at Apache, I think we stand a good chance of succeeding.
|
