The PeopleAggregator is starting to come together very nicely. Feels simple, usage nearly obvious. A pleasant experience. Pretty easy to define relationships.

And it's built on xml.

Here's my public page and the RDF underneath it.

Some thoughts on relationship vectors...  

When I declare my connection to someonein PA, I pick one of these types:

• know of
• don't know but want to
• know of in passing
• know by reputation
• acquaintance
• friend
• close friend
• relative

The degrees of friendship are very nice and upbeat, but I think it conflates (a word I don't get to use very often) three dimensions into one:

• time (past, present, possible future relationship),
• strength (distant to close relationship), and
• attraction (love, neutral, hate).

I should be able to declare that I mildly dislike Amy, my current wife. Or that I intensely hate Bob, my former boss. Or that I have a crush on Cat, this person I barely know. If the scales are quantified, you can do marvelous things with recommendations, matching, etc.

Add an "other relationship" category to accomodate the many ways we define our relationships. "Slept with", "worked with", "screwed over", "blogrolled but never met", "makes me gag", "lust after", "we're both Elks", "divorced me", and of course "know, but don't want to" (needing to divest). "Family" is also a pretty broad and shallow bucket; Genealogy XML covers more of that ground. And this doesn't even get into culture-specific notions; tribal affiliation means different things if you are a Native American, a Jew, a Boy Scout, or a Kurd. Leave room for those wonderful connections and the results will astonish. 

[a klog apart]