Sunday, October 30, 2011

Intertubes refreshed...

I've switched hosting plans at 1and1 from Windows to Linux, and in so doing I lost a lot of customization related to domains I have for various ideas. is back on the air, though there isn't much there right now...  here's the latest writeup which I posted in a comment to a story by Dave Winer over at Scripting News.

A while ago, I came up with the idea of a "tube" of stuff (at the time the internet was being described as a "set of tubes") and thus begat

The basic idea is to have a metadata file which keeps track of the files that belong to it, and synchronizes those changes to / from other copies of the tube. Example use cases include being able to share all of my photos (currently 336,637 files in 460 Gigabytes), in a sane manner... like this...

I give someone a copy of the tube containing all of my photos, it's basically just a metafile, taking a few k to hold the root information, and a revocable key which allows the holder of that copy to access some or all of the original tube. This gives them access via an api (hand waving here...) to get a list of all the files, and pointers to resources such as a thumbnail connector which can generate arbitrary sized thumbnails and send those instead of the originals.

Once the person I've given a copy of my tube has it, they can start to browse, at first they just see the list of folders, then they could start to dig (assuming an internet connection on both ends)... they would see thumbnails once they went into a folder (retrieved using their copy of the tube's key), which would then be stored locally, or however their policy decided to handle it.  If they liked a thumbnail, they could get larger sizes, or the original.

Once they actually view the photo, they might wish to tag it, add notes, etc.... this new metadata would be synced back to the original tube, if they had the proper permissions. (why not, it can be revoked?)

Thus, I could share all my photos, and not have to give someone a 500GB drive and pray they were willing to wade through it.

This concept of a shared metadata pool extends the idea of RSS (as I understand it) to be a two way communicating container of objects AND their metadata. (The metadata is the valuable part that always gets lost when you have to push single files to people)

This could work for any number of types of files, video, music, documents, etc. It would just be some text files and API calls, it would be totally transport agnostic. Sneaker Net, TCP/IP, HTTP(S), FTP, BitTorrent, whatever works.

I think that you, Dave, have the wisdom to push something like this past the crazy idea stage towards something that people could actually implement and use... I'll keep the idea alive as much as I can as well.

What do you think?

No comments: