There are issues with both and we are trying to avoid writing our own. > > JsonML claims to be a round-trip transformation, but it is not. In
particular, it does not have any way to represent the XML
declaration, comments, processing instructions, or material outside
the document element. Furthermore, CDATA sections, entity
references, and character references cannot be distinguished from
their literal equivalents. But if these don't matter to you, it's a
pretty good mapping.
How does it deal with XML-->JSON for deep mixed content? I have TEI
documents with analytic markup, on which the publisher's rendering and
the user's linguistic or semantic analysis depends, nested up to 10
levels deep. I am under pressure from some quarters to "make the markup
simpler" by using JSON.
