First of all I'm not at all sure categories have a place in the
protocol at all, or for that matter anywhere in core Atom.
Outside of entries/feeds, they seem like the next most important thing
to get interoperation for.
I only shy away from allowing multiple parents because it gets more
complicated at that point. Do we select one parent as the "primary"
parent, in order to provide a canonical path from root to leaf? Can we
allow cycles as long as there is a simple path to the root? I do worry
about modeling this too specifically and being too restrictive in the
long run.
If trees aren't general enough, then maybe the simple solution is not
to give them structure in the core, but just to give entries a category
element and let extensions worry about structure. In that case, I'm
still interested in a discoverable list of categories, although
certainly some tools won't want to provide that.
On Sep 15, 2004, at 9:34 AM, Danny Ayers wrote:
Categorization on the web is a hard problem with different solutions
being more appropriate in some circumstances than others.
Agreed; is there a loose element that we can provide that will allow
different solutions to be applied via extensions? Can you give some
idea how broadly we should be thinking?