Robert Sayre wrote:
Tim Bray wrote:
I just read this, and it's moving toward being a usable Pace, but it isn't one yet. Could Henry or someone please update it so it has *specific* suggestions for language changes to the draft, then we can see if it gets consensus support.
I agree, but one point mentioned there is bothering me.
"Atom attributes are not in the atom namespace. It is deprecated in rdf for this to be the case. (so I suppose this means it is still possible)."
What course of action can we take with that?
1. specify that processors should interpret unprefixed attributes in atom namespaced elements to be in the atom namespace. This is idiom as specification.
2. specify that all our attributes are to be prefixed. This is specification as overkill.
3. mutter darkly about that damnable namespaces spec, and do nothing :)
I've seen heated arguments over 1 and 2. The argument for 1 is that you only need prefixes for cases where you're dropping attributes into other people's elements, for your own elements prefixes are effectively redundant. The argument for 2 is based on the fact that there's a hole in the ns spec, screw idiom or effective redundancy.
RDF requires XML artefacts are mapped onto URIs; it's much easier to do that if you know what the full QNames are to begin with.
My preference is for 2.
cheers Bill
