It's useful to be able to inspect HTML DTDs without the need for a DTD
parser. So I've created these JSON format representations of the HTML
4.01 DTDs. The process was:

 1. Use dtdparse to convert the DTDs into an XML format.

dtdparse --declaration sgmldecl --nounexpanded <DTD> > <OUT>.dtd.xml

 2. Write a Ruby script that uses REXML to tear out just enough of the
    structure of the output of dtdparse to capture all of the information.
    This builds an object hierarchy that is serialised to JSON.

  Requires: rexml, json.

Please note that the Ruby script has only been tested with these DTDs as
inputs. It won't work for other files produced from dtdparse without some
hacking.

Also, the JSON files aren't exact representations of the originals. I
haven't included all of the entity references in the output, for example.
If you need more you should be able to add this in with minimal effort.
See the source for details.

Send me your patches if you extend the script to be more complete!

If you want to build the JSON files yourself you'll need to grab the DTD
files and the entity reference files (.ent) from w3.org. The sgmldecl is
also on their site.

Also! Frameset.dtd won't work without a small modification. You need to
make the reference to the transitional (loose) DTD obvious to dtdparse:

<!ENTITY % HTML4.dtd PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "loose.dtd">

--
Dom Marks