On Oct 30, 2011, at 11:09 PM, Radim Kolar wrote:
>
>>
>> Obviously Oracle is going to use Eclipselink/Oracle Toplink for Glassfish and Weblogic as they have investments in those projects.
> Oracle donated toplink to Eclipse because nobody wanted it. Because Hibernate is owned by jboss and openjpa/geronimo is owned by IBM then Oracle also wanted own JPA.
No one forced the Eclipse Foundation to take Toplink/Eclipselink, if its sucks as hard are you are implying then they would have used something else.
>
>> It took OpenJPA 5+ months to release a JPA-2 (JSR 317) complaint version.
>> It took Hibernate ~5 months to release its JPA2 compliant version as well.
> It does not matter. J2EE is very slow moving market. How many years after releasing J2EE 6 specs people accepted J2EE 6 as standard for new applications? And remember complains about ending free Java5 support. Today Java7 is out and Java 1.4 is still used in production.
Again, this all depends on the customers requirements. Maybe J2EE is slow moving in your market but its not in mine. We were using EE6 then same year it came out, just because your customers are stuck in the past with tools that force then to work harder, not smarter doesn't mean the rest of us should suffer.
There's still people using Windows 2000 and Lotus Notes. That doesn't make them "industry standard", certainly not in my industry.
>
>
>> Each impl. has their own strengths and weaknesses. I've seen studies that show Eclipselink beating the others in certain tasks and not in others. It depends on what you application needs, just pick the the impl. that fits your applications needs the best.
> I am not going to buy that. If you look at jpa bench suite than both hibernate and openjpa executed it without errors while it failed in some tasks for eclipselink. Eclipselink is really faster then openjpa but for me stability is far more important.
I'll assume you are referencing JPAB [1]. This was last updated in 2010 and in the case of Eclipselink uses 2.1.1 so these results are out of date. This study doesn't include Oracle as a DB paired with each provider tested. Seems like an important DB to leave out of the test matrix. I wonder why. This study is sponsored by the people who make ObjectDB and ObjectDB blows all the other providers out of the water in it. Just sayin. Why aren't you advocating ObjectDB be the 'industry standard'?
This test uses JDK 6u13 on XP Pro SP3. XP Pro SP3 is deprecated by Microsoft.
My point is that I would not draw a conclusion that something should be the 'industry standard' based on one study that was done using old versions of the implementations in questions, on a deprecated OS, using an out of date JDK. The errors you make note of may not even be present in current versions. They may be a flaw in the test, who knows, its all not very relevant at this day and time.
>
> But i dont think that average glassfish user will care about openjpa vs eclipse link as long as he does not have to pay oracle for weblogic.
Glassfish is the EE RI. It makes no sense to bundle as default a non RI spec provider in your EE RI implementation. If you want OpenJPA in GF you can add it. Same situation with all the AS vendors if you don't like their default implementations.
I'll say it again, all the implementations have their pluses and minuses, pick one that works for you.
-Noah
[1] -
http://www.jpab.org