Daniel Sabbah, CTO of IBM Software, gave the keynote address at AOSD 2004 today. IBM officially backed AOP, even going as far as saying that it is "critical for our survival". Daniel came forward with information detailing the fact that IBM is using AOP in a big way right now, and how they are investing heavily in its growth.
- Posted by: Dion Almaer
- Posted on: March 24 2004 19:33 EST
These are very bold statements to come out of IBM in an official capacity.
Ron Bodkin organized a panel on enterprise AOP. The panel consisted of AOP tool vendors such as Bill Burke (JBoss), Jonas Boner (AspectWerkz), and Adrian Colyer (AspectJ). There were also industry practitioners such as Jon Tirsen (ThoughtWorks), and representing academia, Gregor Kiczales ("father" or AOP).
Read todays coverage of the AOSD 2004 conference
Also read about how "IBM plans to bring a software development technique that has been the subject of theoretical work for years to commercial products this year and next." in IBM touts new 'aspect' for software coding
- IBM endorses AOP: "It is vital for our survival" by Ved Gangaputra on March 25 2004 00:41 EST
- IBM endorses AOP: "It is vital for our survival" by johnyzee on March 25 2004 04:54 EST
- It's still too early to say .. by asdf asfda on April 01 2004 13:26 EST
Good to know that the big vendors are embracing AOP !
Perhaps not surprisingly, their first major success with AOP is in testing/analysis. So far, AspectJ is well positioned to be the ultimate testing language. I guess we'll have to see whether or not they decide to start using it for production runtime applications.
Both Danny and Adrian made it clear that they are moving beyond development support aspects. The common experience at IBM (as at other early adopters like Check Free and our customers) is that they start small and use aspects to achieve benefits in areas like tracing, testing and capturing diagnostic information about errors (what objects ran not just a stack trace). They then move on to build them in to infrastructure and business logic aspects.
Both Danny and Adrian made it clear that they are moving beyond development support aspects. The common experience at IBM (as at other early adopters like Check Free and our customers) is that they start small and use aspects to achieve benefits in areas like tracing, testing and capturing diagnostic information about errors (what objects ran not just a stack trace). They then move on to build them in to infrastructure and business logic aspects.We did it the other way round. We started from scratch, with almost no money, no customers, no AOP framework, in one of the toughest segments in computer software (CMS market, there are at least 200 competitors in Sweden alone), and decided to build the software using *only* AOP and with our own framework. The details of this development are available in my blog.
That worked pretty well as today we have money, lots of customers, a kick-ass AOP framework, and less competitors, and I guess it's kind of ironic that we are juuuust about to kick out WebSphere Portal in a huge customer case. C'est la vie.
What's critical for IBM is a steady stream of new technology that they can hype to sell consultants and software to gullible technology executives.
This is a tad cynical don't you think? :)
I am sitting in a talk given by Adrian (AspectJ) and he is talking about how they went in to a very skeptical WebSphere group, and took a challenge.
The challenge was to be able to take EJB support out of WebSphere at build time. This is something that the WebSphere team knew was tough, and if AOP could help they would sit up and listen.
This is all about technology. AOP for the sake of it means nothing... however people are finding that it can save them time and money.
Anyone that has ever wasted time trying to make WebSphere CMP work (not just to work efficiently, but to work at all) will be very glad if IBM tries in the future an approach similar to JBoss (based on AOP and proxies, as far as I know).
IBM's CMP container (with its pre-historic EJBDEPLOY with MS-like error messages tool) is the worst piece of software I have ever seen. Even IBM hired support can't make it work at all. We are considering migrating to Hibernate, since at least it works (and is portable). Too bad, because CMP's are a nice idea, and - when **properly tuned** - can be very effective (as they are in JBoss and WebLogic).
JohhnyZ is exactly correct on this one. IBM is a giant marketing hype machine.
AOP is still at its initial stages, I think there are some creative concepts in it, but's it's too early to say AOP will replace OO or something like that. It's better to see it as a new technology for framework, or a new language like perl, java, .etc, the theory dosn't changed much.