The idea behind troken is a kind of delicious for the desktop. Directories are important notions in file systems because they demarcate boundaries. For example the cairo source code is encapulated in a directory tree that provides some rudimentary organisation over the structure.
For latex also one needs a directory containing the source of a paper. The various figures etc.
One needs facets however and one wants to inherit facets from directories and from activities. E.g. if I make a directory as realm::academic, then I want all files in that directory to inherit the tag realm::academic. Some tags can be derived from the file name, e.g. paper.tex can inherit the tag file-type::latex.
When retrieving files there are defaults that one wants the computer to know about, e.g. these are my favorite paths through the facet space, e.g. realm::academic, realm::personal, file-type::latex
Another tag could be realm::system to denote files that are managed by dpkg.
One wants the tagging system to automatically tag files, to track their movements through the system, and to observe the users interaction with the files. So for example if I tag my current context as search-topic::sat-solvers then I want to mark all pages visited in this mode as relevent to the search topic, or perhaps have a quick button that marks them.
These java script buttons make interaction with the web browser much easier especially since the troken server can be running locally and listening to user input.
So there are three types of tag inference that one would like to have:
1. observation of user behaviour 2. dependence and exclusion between tags 3. agents (computer programs) that can classify objects
based on their user name, content or other accessible features.
One thing that would be nice is to be able to just dump files in the troken directory and mark them up with a few tags.