Continuing with the theme of T-shirts, I'd like to let you know "No Data No Fun" is a cool line printed on my T-shirt I got at a Talend R&D summit organized at a second-to-none level back in early October. I guess having a collection of good T-Shirts is one of the real perks of the developers involved into the open source development :-)
"No Data No Fun" is also one of the themes behind Talend's continued investment into the tooling which facilitates the interaction with Big Data ecosystems. Getting such a tooling done right is hard. I'm impressed seeing companies like Lenovo liking it.
From my point of view, I'm interested to see how an apparent gap between the world of a typical HTTP service application and that of a Big Data one can be bridged. Ultimately web applications are about exploring the data and feeding them back to the users. We've done the first baby step, provided a FIQL to HBase query client that can be used to query massive amounts of data from HBase databases. JAX-RS StreamingOutput would very neatly fit in there.
However, it is also interesting to see how CXF services can be run natively in Hadoop, to save on a data delivery from HBase or other Hadoop-bound database to a query client running in scope of the CXF server, much cheaper to get it straight from Hadoop and send it back immediately. This is something I'm hoping to find some time for investigating next year. Propagating Kerberos or OAuth2 tokens into Hadoop/etc is also of interest.
I hope CXF will help you get a lot of data from Hadoop and have a lot of fun along the way :-)