I like the High Scalability blog but a recent post on the death of middleware did leave me shaking my head. Although discussing the idea of Platform as a Service, the article argument argued that the traditional middleware layer is dead, replaced by items such as RabbitMQ or ActiveMQ.
I can see the argument if an application purely needs to communicate with a single or homogenous data source through the publish/subscribe model. A simple pushing through these does not need middleware since there is a simple workflow.
This appears to miss the notion that, if building a platform which integrates data from differing services and formats, you need middleware. If integrating data sources, then you cannot get messaging systems to push the data out and then ask the application layer beneath the view to munge it and then present it. That would make it hideously slow, or the messaging layer sits on top of the middleware layer to push the integrated data or to limit access to certain pieces of data.
In this case, I’m afraid that the post does miss the point in this case. Middleware is not dead in some cases. Arguably these are more enterprise related and solve certain issues. Some of these are not relevant to ‘Web 2.0’ sites, some are. It is up to the developers and architects to make the call but middleware is not dead.