The result of this set of assumptions has been powerful

The good Applications know what they can count on getting, and most frameworks are built around common assumptions
The bad I needed to transform a document containing entities without wrecking the entities. And now all my CDATA sections are mush. This doesn't behave well in my workflow!
The ugly Developers are frequently struck between the "Infoset" approach and raw text-processing, with few tools offering much in between.