• The XML I see in the data world tends to be message-oriented: often a very flat structure, usually much smaller files, relational data wrapped in tags, SOAP messages, WSDL, etc. Developers usually love tools like XML Spy: they vastly prefer schema to DTD, and as structures are often defined automatically they don’t see much value in human-readable schema.
• Document-centric XML has to handle, um, documents, with characteristics such as component reuse, hyperlinking, cross-references; the things DITA handles are generally relevant only to this side of XML. Many in the document-centric world still hand code DTDs; those that understand Schema all prefer RELAX NG to the junk that the W3C came up with.

Of course both are XML and there are exceptions both ways (some messages get as complex as documents, some documents benefit from strong typing or namespaces) so it is a continuum without a strongly identifiable border.

When we started Silicon Publishing in 2000, we stopped using Xyvision. Initially this was not our first choice but economic reality. However, being forced into early adoption of InDesign automation turned out to be a very good thing.

Xyvision was (is?) great! Very fast, great composition, one of the first/best tools to automate typographical craft. I hope SDL keeps it going; they sure have acquired many companies. I still like InDesign Server better as it surpasses the typography/graphics, but I miss the speed.

thanks for a good read