The other day, Chris Aldrich brought up an interesting point in decentralised systems: How to discover each other?

I have some ideas on this (and likely write some code in the future to explore their feasibility).

The easiest one would be: manually.

I believe, since the very beginning, bloggers curated blogrolls to point to blogs they follow. This way you can build a social graph with a certain degree of confidence.

Personally I would combine that with some kind of trust. How favorable is it to follow an other blogger?

You could look at the links she gets. Ideally over time, so you could handle hypes or similiar.

Recently I discovered FoaF. From what I understood it is a file you put on the root on your server and point to via <link rel="meta" type="application/rdf+xml" title="FOAF" href="/foaf.rdf" />.

The structure lookslike this:

  <foaf:PersonalProfileDocument rdf:about="">
    <foaf:maker rdf:resource="#me"/>
    <foaf:primaryTopic rdf:resource="#me"/>
    <admin:generatorAgent rdf:resource=""/>
    <admin:errorReportsTo rdf:resource=""/>
  <foaf:Person rdf:ID="me">
    <foaf:name>Full name</foaf:name>
    <foaf:givenname>First name</foaf:givenname>
    <foaf:family_name>Family name</foaf:family_name>
    <foaf:mbox_sha1sum>obfuscated mail address</foaf:mbox_sha1sum>
    <foaf:homepage rdf:resource="FQDN"/>
    <foaf:depiction rdf:resource="FQDN with path"/>

This code was generated using FOAF-a-matic.

You could list your contacts there, too. I haven't decided on whether I want this, though. My concerns are about privacy and respect:

What does it tell about them and me if they show up there? Could people feel offended if they don't make it on this list? Since you have to pick a predicate to describe the relationship, which one is the most appropiate?

Another approach I discovered is called XFN. It is also XML-based.

By using XML you express the semantics (like "A follows B") in a machine readable way.

This could be used by programs tohelp you discover more people.

What I find more challenging is to parse podcasts for interesting people. This requires manual work from what I can say right now.