How data can learn from documents

XML as a series of collisions 'SGML for the Web' hasn't worked out to look much like the 'traditional' HTTP+HTML Web. A document format has proven more useful to data-oriented developers than either document or Web developers in the traditional sense. Much of XML's success has come from the continuous repurposing of document-oriented technologies for data processing.
XSLT as a start XSLT, despite its obvious origins in document formatting needs, has been enormously popular with data-oriented developers moving information between different data structures. The document-centric nature doesn't make for a perfect fit, but the fit is generally 'good enough'.
Keeping data free of processing Perhaps the fundamental gift of XML has been its demonstration that complex data structures can and should be exchanged without accompanying logic.