Transformations as Profiles

On the 24th I wrote up a huge summary, currently unpublished, of my Semantic Web work to date. This got me thinking about the recent work that I've been doing, and I ended up chatting to Earle Martin and Evan Prodromou on #swig about Earle's vocabularies and how they could be improved, and Evan's informal NID request, which I sent some followup email about. Incidentally, on the email being a mass noun question, John Cowan noted that "mail" is a noun, and "post" for that matter, so why not email?

I chatted to Earle about, for example, what the best way is to model checksums of files in RDF. Of the three approaches—flat, datatyped, and interpretation properties—I've decided so far that datatypes won't work because of the query aspect, and the flat approach is probably best overall. Earle doesn't yet seem to have heeded the advice, which is fine, but nor has he given any indication or argument as to why his overcomplicated approach has any benefit.

After that I chatted with Dan Connolly in the same place about file metadata, microformats, shells, sensible defaults, Mercurial, GRDDL, and some recent data work he's been doing. This eventually led to me posting on miscoranda about GRDDL for XHTML Schemata Associations, which in turn is making me wonder about a RDDL-in-GRDDL approach, and the implications of using transforms as globally unique identifiers. Since, for example, the nature of a GRDDL transformation is currently open ended, RDDL-in-GRDDL could itself be used as a transformation. I'm also starting to think that the TAG's current position on namespaceDocument-8 is a cop-out, and that a specific approach should be chosen rather than a proliferation and proselytisation of the "model". There would be far more utility in a standard approach, just as robots.txt was far more practical than letting each site decide how to express that information.