Andreas Savva wrote:
I don't agree, however that these filesystems should be 'defined and available' by default. I think they should be simply well-known names that can be expected to be (typically) available with the definitions provided. And they should only be made available only if they explicitly appear in a JSDL instance document.
I think the question here is to what extent should we define the "well-known" filesystems. My proposal is probably the limit of this, and I'd agree that being explicit is going to be better practice. In any case, there are going to be important semantics attached to some filesystem names that are going to be tricky to capture (e.g. why should HOME refer to the user's home direcory?) We may well be able to punt on this matter until post v1. A separate matter is whether filesystemName references should really be xsd:QName instead of xsd:NCName (if so, the attribute should possibly be called filesystem), but I'll let the schema hackers decide what to do there. I'm more concerned about the mechanism, and not the syntax for expressing it. :^) Donal.