Thursday, September 02, 2004

Infomerge a go-go

Finally gotten around to reading the Networked Semantic Desktop paper, which has been sitting open on my desktop for a week now, and I like what it sayeth. Anything with a heading of "Surviving the Information Flood and Creating Knowledge in the Process" has got to be good...

It doesn't really suggest anything concrete - it's more an overview of existing efforts in 3 fields - p2p services, social networking, and semantic data - and the combination and future of all 3 of these to create distributed knowledge that's accessible by both machine and mind. There's a fair amount of talk about how metadata can emerge from social networks, which has got me thinking a fair bit.

Currently, I think it'd be great to be able to use the existing p2p networks for more social-network-based activities. For instance, imagine creating an RSS feed of "documents" you're currently reading (including books, PDFs, web pages, e-mails, etc.) and being able to subscribe to other's similar feeds. This would form a kind of p2p network which was then backed up by something similar to existing networks - distributed file storage - with peer-based reputation to advise you on which documents you may find most interesting.

The problem as I see it is that the current range of p2p networks are to specified, and choose to concentrate on either one type of file (such as music), or one type of usage. At leats with the network I use most (eMule/eDonkey), metadata is only concerned with the format of the file, not its semantics. You can add comments to files, but I don't think you can search through them. There needs to be a further level of metadata for *human* use. It's also intended for more "public" networking, rather than more social/reputational peering, and so the queuing system implemented would perhaps get in the way of usable document peering.

I hope to follow up the other software and networks cited in the document, but some of me wonders if there's too much diversification going on here. We either need something so simple that anyone can take it, use it and adopt it to their needs - if this doesn't exist now, I suspect it's not long until it does. - or we need technology that allows people to set up their own decentralised nodes easily, based on existing common protocols such as ftp/scp/http. This is a form of limiting factor when it comes to any of this decentralised peer stuff - take-up and economies of scale.

I shall continue my search as time allows...

No comments: