I've been pondering the problems of creäting decentralized equivalents to social network sites such as LiveJournal.

A 'blog per se comes fairly cheap. At the low end, one could form or join a syndicate, jointly register a second-level domain name (eg, oursyndicate.com) with GoDaddy for about $10 per year, jointly lease 500 GB of housing from AN Hosting for about $85 per year, distribute third-level domains (eg, winky.oursyndicate.com) amongst as many syndicate members as you might have, and install WordPress for free. Give everybody a whole gigabyte, and we're still talking just 20¢ per person per year. (Let everyone have his-or-her own second-level domain, and 25 GB, and we're talking about $14.25 per person per year.)

The challenge is in giving such 'blogs — across second- and third-level domains — the connectivity of Friendships, and of Interests.

As a first-pass approximation, imagine each 'blog as having a link that will deliver two data: an OpenID associated with the 'blog, and a reference (pointer) to an RSS feed for the 'blog. These data allow one to distinguish the 'blog owner, and in some sense beFriend him or her.

The next stage is to support a Friends page. A question is of whether to just deliver a set of references (presumably URIs) for the 'blogs of one's Friends, and leave aggregation to the visitor's software, or to assemble a Friends page at one's own site. The advantage of the latter is that the visitor needn't have aggregation software; the disadvantage is that either the page will have to be aggregated on-the-fly, or it must always exclude protected entries. I'm inclined to opt for aggregation on one's own site and for aggregation on-the-fly. However, a standard could support all of these options, leaving it to a given 'blog to decide whether to deliver just references, just Friends pages, or both, an whether any Friends page were aggregated on-the-fly. The next piece is Interests. A perfect decentralization of these is possible but (as I think) not reasonable; it would involve shipping copies of a large dB repeatedly to each 'blog. My thinking is that, instead, we should accept there being Interests servers, at which 'blog owners could register their 'blogs as corresponding to given Interests. However, no 'blog need be dependent upon just one particular Interests server; it should be possible to register with multiple servers, so that any Interest-search censorship on the part of one server could be overcome by merely additionally registering with another server that did not censor that Interest. It might be possible for these servers to be supported charitably; but, frankly, I imagine them supported by advertising or by registration fees.