Some moons ago I wanted to do a similar thing that I ended up not doing, and that was parsing tags that including spaces. I didn't get past what syntax I'd use but with all the tagging a lot of sites do now I am sure this is a solved problem. Additionally, I was looking for the flexibility to associate an "attribute" with a tag or set of tags. The application I was writing would have benefitted greatly from this sort of scheme.

For example, "tags"

• word1
• "word2Tag1 word3Tag1 word4Tag3"
• ATTRIBUTE:word5
• ATTRIBUTE:"word6Tag2 word7Tag2 word8Tag2"

Inline, a real example would be something like:

humans
"good picture for thumbnail preview"
boysEyes:brown
fathersHair:blonde
description:"a father and son are playing catch"
[download]

This is obviously not a solution but I wanted to expand a bit on the scope of the _problem_ space. Maybe some wisdom will enlighten how to approach this since I am still interested in figuring this out since it'd have applications for labeling things like images for the borg. At this point there might be a well established way to do this that is not SGML/tag-based.