Mihai Lihatchi

I had passed the SCEA 5 beta and it was no where near that difficult. At least for me it was straightforward (without having any preparations) and the passing score was lower (60% I believe) than for this certification and the questions .. a lot easier ( no Adam Bien patterns , no Bitter EJB patterns just plain old CoreJ2EE and GoF + they were never that ambiguous).

I did not invest more time or money back then either. Just applied for it 2 weeks in advance.
Both times I was employed and busy. Passed with a comfortable margin then .. now barely made it.
The SCEA 5 beta counted towards my architect certification from Sun.
I wonder if I can get away with OCMJEA 6 without also doing the part 2/part 3 and training parts if I passed the beta (equivalent of upgrade exam ).
Asked Oracle but nobody answered .. yet.

Got mine .
Barely passed .
I won't go any further since I am already doing architect work and it is really tricky if you ask me to answer the questions.
It is true I did not prepare long enough .. just 2 weeks after work but still .. I wish I could be more confident when answering the questions.

So today I also took the test.
I will confirm the questions as in the objectives PLUS : Core J2EE Patterns (some of them were removed with the new Java EE 5).

Lots of patterns from GoF (some not mentioned in the objectives).
Also references to LeakyAbstraction and HotPotato
Webservices as mentioned above :JAX-RS vs. JAX-WS combined with security/asynchronous approach. In general the integration questions were also the most difficult during the other SCEA Part 1 (also beta).
Also references to singleton session beans and comparison with stateful session beans -so EJB3.1 is on the exam clearly.
Two questions about SOA (hmm zero experience there for me so I chose some answers that looked right).
At some points I felt the exam was getting too much into details of JPA (but since I only have experience with the Hibernate implementation it looked to me that way).
Regarding web-layer : Some questions involving JSF. My first instinct is to REJECT JSF but some usecases just demand it.
The interesting part was about deciding on hardware for the platform .. that was mostly a managerial issue (time/costs constraints) and I tried to provide a good answer but I never was in such a position.
My feeling is the exam is more difficult than SCEA 5 PART 1.
Although there was plenty of time I bearly made the second pass properly. Some questions seemed to lack a second option and were a huge time drain (maybe this is what you found ambiguous too).
Preparation time : 2 weeks part-time. Books : Mostly the ones for SCEA-5 plus reading on anti-patterns and Adam Biens presentations. I have a copy of "Don't make me think" but to handle just one tiny objective about ui design principles it seemed like an overkill.
Should be one of the last to take the beta so hopefully this post is relevant for part 1 (real one)

Well anyhow since there is still a training course involved I do not think I will try again if I fail this one . Anybody knows about the "upgrade path" to SCEA/OCMJEA 6 ?

Well I got my SCEA 5 in 2010 just before the Oracle changes to mandatory training
It is probably not 100% correct to require a training for people that can pass the certification. The training should provide real value for the future Java architect by itself.
Although few people actually checked whether I was indeed Sun Certified Enterprise Architect or not .. most just did their own technical tests during the interviews.
So it is fair to declare you passed the requirements for Oracle Certified Master Enterprise Architect .. not that you are . Maybe it is a lie by omission but in this case it sounds fair and you can explain that the training did not provide value.

Time for my 2 cents .
It does bother me but if you have a look at the code in the sample Depot Application for the models (generated by Netbeans) you can put nice annotated docs to suggest the fields (I believe the book "Agile Development with Rails" specifies more).
I got used to having the DB console open in the first stages of developing the application. Of course with Java/Hibernate things are not so bad but it's not a major block.
The fact that you can discover the error in an instant kind of compensates for this small annoyance.
This reminds me I should use that plugin for my current RoR code .

Well since I am such a noob .. I figure I might answer this.
All you need to do in these cases is to restart the development server (WebRick). Meanwhile I have managed to get my application going at a pretty good speed and I think that Ruby on Rails is great for small web applications.
Unfortunately I will stick to J2EE for really critical enterprise apps. Things like 2 phase commits and lack of standards deter me from considering RoR for more elaborate enterprise solutions .. for now.
But as I am sure most of you found out too it bring back the joy of programming a bit.

I wrote a longer entry on my blog (http://lonelywolfsden.blogspot.com/2010/12/ruby-on-rails-or-beauty-of-instant.html) but as a Java developer I am positively impressed with the possibilities opened by Ruby on Rails.
I was trying to implement a complicated GWT +MAVEN + GAE application for a month and got lost in some configuration details. . so I just left the project in pieces on my SVN folder. A week before I had installed the sample depot application of Rails in Netbeans. Saturday morning I had started to create a new RoR application.
After just 24h I had managed to get something working and using it (personal pet project).
The key features of RoR for me :
1) Scaffolding
2) Instant feedback ( according to http://www.zeroturnaround.com/blog/java-ee-container-redeploy-restart-turnaround-report/ we loose 25% of our time just compiling and deploying ) which I think is crucial these days.
As an ex-PHP programmer I can say that Ruby(on Rails) has the power of Java (and the good software engineering approach) with the instant developer feedback of PHP (interpreted language development).
Would I architect a bank application with Ruby on Rails ? No .. I mean maybe the front end (web layer with JRuby at most).
Would I do anyhing I implemented in PHP before in Ruby ? Of course .. 1000 .times { puts 'yes'}
I am dissapointed that people on Javaranch do not write more on this forum but nevertheless I am happy to be writing here.

Now to ask a question : say I have created a model with a integer value like car.max_speed. How do I change it to float without regenerating the model ? The field in the DB is ok.

To the OP.
It took me 2 months to get a PHP/Perl/SQL/Java coding job when I finished the University where I had coded a lot of Java and C/C++.
The first professional working code was in PHP and in PERL so I can't say I did not have to downshift a bit and to demand less from my 1st job.
So you need a basic coding job which means a "junior" Java developer or even PHP job should help you for now.
Don't be afraid to jump back in the Java world when and if you need to (I did and it was fine).
Also remember that being a good software engineer means more than just Java ,it means having problem solving skills and business understanding skills.
I believe once you know a language like C/C++ very well and you have the other qualities you can learn Java in a few months and be good at it.
So tell the prospective employers : I will need 3-4 months to get up to speed wth Java but I will be able to be productive after them , in return I am willing to work for a lower pay. Most employers I know appreciate somebody who is fair about their skills and willing to improve (and asking a lower salary is just the cherry on the topping).
Good luck !

I noticed (a bit annoyed) that the latest version of Netbeans (6.9.1) does not offer the UML plugin offered before http://netbeans.org/features/uml/
Can't help thinking that this event is connected to the Oracle control over Java and Netbeans. I am a bit disappointed as I never expected this to happen even though Oracle is known for it's marketing techniques.
I happen to have a Netbeans 6.7 at work which has the plugin and it is quite useful but it's not at work that I need it (I have some work provided tool there), it's at home on my Ubuntu Linux machine for doing my own personal/hobby projects.
So you Linux software architects/engineers what do you use ? I used Umbrello but compared to Visual Paradigm/StarUml it is just not at the same level.

Congratulations Anurag. Your work finally paid off.

Well guess what :
http://techcrunch.com/2010/10/09/google-car/

Google is on to the race and it looks like in 8 years (say a decade) we will have a working AI driven car.
Of course they wouldn't have done it without Stanford's AI Lab and it's not clear to me whether it's Stanford's Audi TT (well VW's actually) or Google's own Audi .
Anyway I am hapy they are working on this and hope I will not have to renew my driving license (expires in 2018 ).

Umbrello is a nice Linux tool you can use. However for the purpose of doing the diagrams I would recommend StarUml even if this means to have a temp windows install (use a virtual machine if you wish). I could use the Windows machine at work (after work hours) in order to play with Windows tools. See if you can get your hands on yours .

Of course I used a tool and you need a free one in order to avoid having watermarks in your diagrams. Or to pay a premium price for a commercial tool.
My handwriting skills are not brilliant so .. I dismissed the development by hand.
That being said the requirements are to have a set of diagrams (how are they generated is irrelevant) placed in some HTML files.
If you think you can do it without a major hassle you can always draw the diagrams by hand and scan them when you are done .. but I think it is not very practical when you need to modify.
At work I often draw the diagrams on paper for some simpler problems.
The famous Dijkstra (http://en.wikipedia.org/wiki/Edsger_W._Dijkstra) used to avoid computers as much as possible which now sounds weird considering his contribution to computer science .

Hello Luke.
Using a JEE cluster is the recommended way to deal with the single point of failure problem (and to improve scalability, overall performance)
Somebody posted links/info on clustering on this forum (if you search closely)
Regarding collocation (deploying web and EJB packages in the same server) it depends on the requirements of your system :
If you only have web clients I would go for a collocated solution for instance (about 90% of the cases) , if you have other clients (like a Swing desktop client, WS client) then consider a distributed ejb and distributed war solution.

Anyway this is just my opinion you can do some more research and post back .
As an ex-Spring developer I am a bit biased against the J2EE distribution (as in separate servers for EJB's and WAR's) but there are definitely cases when it makes sense do don't dismiss the solution.

I did not mean to say that but I know that behind the wheel people change their personalities a lot ... and not for the best. And like it or not a computer is not likely to take a phone call while driving at 100 km/h in heavy traffic. Some people will ... some won't , some oare overconfident in their driving abilities some are too scared.

Regarding trains I know for sure that in France at the Charles de Gaule Airport they have A.I. controlled trains between the two terminals.
Didn't a human error caused the train crash which occurred recently in Washington ? This year if I remember correctly ?
The bottom line for me :
More cycling == less blood pressure . But today it rained maybe I see a velomobile in my future.