You may recall all those discussions about "RESTful services can be consumed from browsers by humans only" statements. It was awhile back and of course it's been proven since then that it's possible to write a sophisticated client code consuming RESTful services in a number of ways.
In JAX-RS land, Paul Santoz innovated with introducing the fluent API with all/most JAX-RS stacks having custom implementations, and this API is now being standardized by JSR-339 expert group. Proxy based API is also supported by some stacks.
What is important to realize now is that JAX-RS client runtimes need to become more robust and sophisticated in order to move further, beyond supporting the assertion that "yes, we can do it, we can write the client code for working with RESTful services".
A human working with the browser has more space as far as time and decision making is concerned, when facing a connection failure for example. On the other hand, the code needs to be smarter and ready if the expectations that it does not exit immediately after a connection has become slow, broken or a target endpoint has been recycled is to be met.
This is where a failover feature comes in and it's been supported for CXF JAX-WS clients for a long time. Starting from CXF 2.4.1 it's also the case for CXF JAX-RS clients.
Please check this page for more info, experiment and provide the feedback.
It also confirms that in CXF we are committed to bringing the best support for developing SOAP and REST services. SOAP and REST developers have their 'differences' :-) but we understand them, bring both 'camps' together, and do our best to support them at the framework level.
As a side note, if you think about it, making clients failover-capable is a cheap way toward ensuring something close to 24-7 up time. If you are Amazon or BBC then you have all the hardware and experience in place to ensure the clients nearly never experience a "can't connect" problem. What if you are running an OSGI container hosting service endpoints and need to replace a service ?
Failover-enabling clients is a simple way toward making service endpoints replaceable without affecting the critical client code being executed somewhere. Configuring a CXF failover feature with a small timeout between retries will do the trick really well.
One thing which is really exciting for me is that my Talend colleagues are already working on providing enterprise-level failover-like features for CXF JAX-RS clients and I'm going to talk about it in one of the future posts.
If you have already integrated CXF JAX-RS in your higher-level product then I'd encourage you to follow Talend's lead.