Although I travel in the wide open realm of ideas, I am not into impractical ideas no matter how attractive they may be.  My excuse for wasting my time with impractical ideas is that they tend to be stepstones to practical ones.  Anyhow, here is one I thought could be of immediate use to the blogosphere.

The idea is to turn all those news aggregator clients out there into news distributors.  So all the RSS files I download every morning is also made available to others.  Whether sharing of RSS files and resources is done through a variation of BitTorrents, blogosphere-specific P2P network, or existing P2P networks is irrelevant as long as the sharing is done by those who can.

I know this is probably not a new idea, but I am peddling it here because I believe this must be done if blogosphere is to continue growing.  At some point, distribution of load across news aggregators must be discussed but that's a post topic for the future.  For now, enthusiasm for something new is enough to power this effort.

My recent burst of interest in P2P networks stems from the idea of sharing wishlists which I mentioned in my comment about CleverCactus Share.  Diego expressed some enthusiasm about the idea so there is a good chance it will appear in ClearCactus Share.

But my mind is already leaping forward exploring generalizations and mutations of the basic idea of sharing and seemlessly integrating the capability into the operating system.  To power this exploration, I am asking questions like:

• What is a file?
• What does it mean to share?
• How can I share something without having it?
• How can I share nameless, locationless, temporary, and formless objects?
• What if everything is shared by default?
• What if bots are loosened into the shared space?

Many of the answers seem to point to a need to review some basic assumptions pounded into us by file systems and databases like names, locations, queries, etc.

If you drag a file from here to there, people expect the object to move from here to there and also expect the operation to take some time.

If the object is a wish, meaning it will arrive sometime in the future, dragging it from here to there could mean it should be moved there when it arrives.  That's like an instruction.  But people usually don't give instructions to empty air.  Is it better to introduce a bot-like objects or should future actions or 'promises' be turned into an object?

There is also the problem of having too many P2P networks.  They can be abstracted or hidden behind other shapes and forms.  At that point, even services like eBay and Amazon can be thrown into the mesh.  An auction involves someone wanting to sell something they either own or can provide.  Drag and drop this from here to there and eventually a real-world object is 'downloaded' to someone's doorstep.  All this is nice except abstraction and usuability don't often walk hand-in-hand.

I know that I am overstepping the bounds of practicality in many areas, but overreaching is often useful when searching through the solution space.  Anyway, these are kind of things I occupy weekends with these days.

I was looking through MLDonkey wiki when I ran into this technical yet compact explanation of how BitTorrent protocol works.  I already know how BT works but I thought my fellow geeks might find this useful since the diagrams at the official BT site aren't that useful:

It [BitTorrent] divides shared data (a single file or a directory) into pieces, typically of 256 KiB. A SHA-1 checksum is computed for each piece, and used to check the piece has been correctly downloaded. The checksums are stored in a .torrent file, along with filenames. The .torrent file also nominates a tracker, a Web resource that introduces peers to each other. Peers contact each other, learn what pieces they have available, request the rarest (least commonly seen) pieces first, and send requested pieces.

Just in case you are wondering, MLDonkey is a universal client of sort for many P2P networks including FastTrack, eDonkey2000, Gnutella, and Direct Connect.  It supports BitTorrent too but then BitTorrent is not really a network.  While MLDonkey is open source, it's written in Ocaml which is powerful but non-mainstream.