While not an editor, but rather a language, Skribilo (http://www.nongnu.org/skribilo/) can output documents in various formats, including Context and Lout. I have worked a bit on getting better Context output from it and last tinkered with the math output about a year ago. Such a system might form the output engine on which an editor could be built. The same might be said for Pandoc, in which case perhaps one of the existing Haskell editors could be used as the basis for a specialised text processing system. For non-technical documents SiSU (http://www.jus.uio.no/lm/toc.html) offers various output formats, but again it is not an editor.
While your concept is interesting, I'm an Emacs user, and unlikely to switch to anything else.
Cheers, Roger --
Hi Roger,
Thank you for the suggestion. I was first thinking about incrementally
creating a custom format that evolves as features are implemented. And
for translating the custom format into a backend format, I was thinking
of creating files with translations rules for each backend so that
anyone can add support for a new backend or update an existing backend
to add more feature or to make it compatible with a newer version of the
backend, without needing to modify the editor code. A translation rule
is e.g. start_section[title=