Got involved in a discussion of FOAF and what it's for over at : Danny Ayers, Raw Blog : Micromodels
I think David Janes understands that applications need to precede specifications because he started off as a screen-scraper. Someone who sees there's an application for the data *before* there's even a machine readable version of it. And he jumps through hoops to try to get at it.
Danny seems keen to conflate this with a completely different issue, whether the *model* is abstractable away from the *syntax* of the file-format. But that's a side-show. The question is whether the model is embedded in (and nourished and sustained by) a practice humans actually want to engage in. Or is it floating optimistically free of such constraints, waiting for someone to take pity on it.
This is not a question of syntax vs. semantics. It's all about semantics vs. pragmatics.