hadoop-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Steve Loughran <ste...@apache.org>
Subject Re: Hadoop Development and the new Oracle's Plans
Date Thu, 03 Mar 2011 10:43:01 GMT
On 02/03/11 17:05, Marcos Ortiz Valmaseda wrote:
> Regards to all the list.
> I have a doubt in my mind since Oracle announced its new plans for Java.

The ASF has left the JCP due to Sun and then Oracles unwillingness to 
release the Java test kit without imposing freedom of use restrictions 
on Apache source code, despite the JCP rules not permitting this. that 
doesn't mean the ASF is backing off Java development, merely backing off 
attempting to collaborate with others by way of a Standards body that 
has more in common with the peoples congress of the union of soviet 
socialist republics than with a functional democracy. We will work as 
open source projects, with all discussion, code and tests in the open.

> 1- How the new restrictions can affect to the Hadoop Development?

none, Apache code.

Sun/Oracle are free to participate; some of the SunGrid people have been 
involved in the past. Their contributions are still welcome, if 
Hudson/Jenkins is happy with the code and the number of tests they 

> 2- Will Hadoop support the new OpenJDK platform?

OpenJDK 6 is effectively Sun JDK with a different rendering engine, the 
closed source one has C/C++ source from things like OSF/Motif and the 
like in there so it's copyright is dirty.

I don't see anyone rushing to move to Java7 for Hadoop in production, 
either in source code or binary. Everyone with large clusters likes 
stable versions, and tends to be trailing edge with Java 6 versions, 
avoiding new features like compressed oops as you can predict that with 
500+ 12-core servers, if there is a race condition in the JDK, you will 
find it.

The other issue is that a lot of people out there use Mac laptops for 
their development, and until java7 ships for the mac, nobody who has a 
mac will be able to develop any Java 7 code. That's another reason for 
Hadoop staying on Java6

What could be interesting would be for Hadoop to have more scala 
integration. I think there are some good arguments for us moving some of 
the high level code to that language, within the JVM, rather than follow 
the oracle roadmap.

Speaking of Oracle Roadmaps, they keep talking about Java EE moving to 
be cloud computing. Well, the Java Cloud Computing API is the ASF stack, 
from libraries like Whirr to talk to the infrastructure, the Hadoop DFS 
and MR APIs to talk to the filesystem and MR engine, and the layers on 
top. They can lay down some standards in the JCP, but out here you get 
to build things that are useful today.


View raw message