I am _so_ tempted to build this
Mark Pilgrim:
The next logical step after that, of course, is to automate the process entirely. Don't like RSS? Change it! Why don't you set up a web form where users can check off the elements they want, arbitrarily misspell some elements (like you did when you stole RSS from Netscape), add or remove allowed HTML entities (like you did when you stole RSS from Netscape), and add a bunch of meaningless illdefined elements like ttl or guid. You can call it the "RSS Spec-o-matic".
Then we can set up a directory of all the world's incompatible RSS specs (as if we didn't have enough already). And we can make the directory available in OPML! Bonus! Then we can all watch the bug reports pour in as users not only don't know which validator to trust, they don't even know which spec to trust.
An automatic spec generator. With OPML output. Heh.
To be truly useful, it would have to generate validators as well. I guess a validator could just read the OPML and work off that.
Seriously, though, it does have a real world use. Recently I wrote some Python code that parses an XML schema (my own proprietary format; not the XML Schema) and generates C code to parse (and validate) it. It lets you do things like bind variables, so you can get it to populate config arrays etc. It wouldn't be hard for me to generate an ultra-strict RSS parser in C with it (not that that would all that useful). The spec-o-matic would be a nice counterpart - an editor for the schema files.