A few days ago Albert found something unclear in the FacetMap model. I
wanted to share my response with the group, since I wasn't sure if it was
different for XFML or just unclear here too.

Essentially, the XFML file had a page with occurrences to different topics
in the same facet. An extreme example:
<facet id="flavor" />
<topic facetid="flavor" id="cherry" />
<topic facetid="flavor" id="chocolate" />
<page title="Ben and Jerry's Chocolate Cherry Garcia Ice Cream">
<occurrence topicid="cherry" />
<occurrence topicid="chocolate" />
</page>

In FacetMap this is illegal. Is it in XFML? My response to Albert went like
this, and I wonder if anyone could offer comment:

Yeah, that's how the model works. The practical reason is, the FacetMap
interface is supposed to let a user very easily and quickly narrow down
many thousands of resources to a selection of just a few. The more topics
you give the user to choose from, the tougher that is. You can see how the
selection would narrow more slowly if there were extra occurrences.

The underlying logistical reason is, facet structures are supposed to
contain mutually exclusive concepts. The XFML spec says this too, though it
may mean something a little different. I mean that a resource should be
described by only one topic in a facet -- and if it is described by two
topics, those topics belong in separate facets because they must describe
fundamentally different characteristics of the resource. What you want to
do is association: "this article is about Grey Matter AND Movable Type" or
"this article mentions Peter Van Dijck AND Simon Willison". Association is
different from classification, and FacetMap is about faceted classification.

Thanks all,
Travis Wilson
Development
http://facetmap.com