Disclaimer: The opinions expressed in this post are, for better or for worse, my own and not intended to reflect the policies or positions of my employer, Oracle, or those of the W3C
If you are looking for the next big thing in middleware and integration, you should look at Linked Data.
In December of 2011, the W3C organized a well-attended workshop on Linked Enterprise Data Patterns. (There is a link to the position papers at the bottom of the page.) The thinking behind this workshop was that although RDF and the use of URIs to identify and link to items of data and REST were gaining popularity they had not yet been applied to solve enterprise-levels problems. Would extensions be required to scale up to the enterprise level? The title of the workshop implied that we needed to develop patterns to apply to particular situations.
Martin Nally of IBM Rational led off the workshop by discussing the need to integrate tools flexibly and efficiently. An example of what he needs in his business is to integrate data about softwate bugs: bug reports, provenance, screen shots, responsibilities and so on. He said that they had tried different technologies for integration (Database integration, Enterprise Systems Bus) but found them rigid and hard to adapt. With RDF and URI integration, you don't need to update the schema to add another related piece of information or access another tool. Here is a picture from Martin's presentation that shows links between different kinds of data.
The same argument can be made for integrating applications. If you have an ordering system and want to add a FedEx tracking link to the customer's online receipt, all you have to do is to tweak the UI and add a button which links you to the FedEx system.
If you use RDF and Linked Data to integrate applications and tools, you need to be able to manipulate the data and add to it, not just read it. This changes the paradigm and enables new kinds of applications. The data needs to be in stable storage with read/write capability. This is a new face for the Web - it makes the Web writable and this means we need to think about the usual "database" capabilities such as access control and transactions.
Based on the success of the LEDP workshop the W3C chartered a Working Group called the Linked Data Platform (LDP) WG. Two of the areas the WG is working on are how to deal with collections and the need for pagination in case the dataset to be presented is very large, questions that REST does not address. Here is the current working draft.
The WG is making good progress and should have a working draft of the specification ready for public review in a few weeks. I will keep you posted.