A while back I wrote out a RELAX NG schema for OPML. It's very simple - it defines all the core constructs in the two OPML specifications (1.0 and 2.0) along with the provided example files. You can use the schema in tools like James Clark's nXML mode for emacs or the excellent oXygen XML editor to write OPML by hand.
The schema is available on a Github repository that I have set up called opml-schema.
I welcome patches to the core schema and to the Schematron rules which exist to supplement the RELAX NG schema. We still have an issue over how to codify extensions to OPML, which are permitted using a mixture of XML Namespaces and the 'type' attribute as per the Extending OPML section in the OPML 2.0 specification.
The question for OPML users, developers and tool vendors is whether this unofficial schema should contain descriptions of third-party extensions. The alternative is we could use something like NRL or NVDL to point people towards namespace-specific schema documents. It's a reasonable solution, but I am not wild about it. RELAX NG tool support is better than NRL or NVDL, including a number of important languages which I am using for my own personal use.
It seems like we can cover third-party extensions in the schema following some rules: the schema describes but does not prescribe; start using first, put it in the schema later. Of course, the schema is open source (GPL) so you can take it and change it yourself. I only have the power to say what goes out in my repository. Git is rather humbling like that. Subject to figuring out the best way of laying out the schema for this purpose, I will probably allow third-party extensions to the schema into the repository - but I do so at my own discretion following the principle of description not prescription and not including things in the schema before they are out there and being used (sorta like microformats!).
I'll also be a lot more likely to include patches and third-party extensions if they are sent to me as a Git patch! If you aren't using Git, you should be. It's simply version control done right. Go watch the Linus Torvalds video.
Anyway, I hope that you like the schema and I especially hope it helps you publish better data.
