Why 'real' XML to the client?

Escaping the HTML straitjacket

Fit for human readers, only While it isn't impossible to reuse information stored in HTML documents, it's far more complex to extract information from HTML documents and analyze or further process it with other tools. HTML doesn't describe information thoroughly, only the document structures which contain it.
Opening gateways An XML-on-the-server approach requires developers to know in advance what client needs will be. Shipping XML makes it possible for clients to make up their own minds how to present the information, potentially transforming it to new formats.
Moving toward automated transactions Distributing XML over the Web reopens some of the failed possibilities of agents processing and exploring information without constant human intervention.
The possibilities of the Semantic Web The W3C envisions the Web evolving into a system of computers where meanings are exchanged, not just documents. HTML was the original inspiration for that dream, but is now more of a roadblock.