Things I consider "a document" and that I would like to manipulate as files in my filesystem:

* A Facebook post.

* An email I receive.

* A form I fill in online.

* A chat session on WhatsApp.

* An SMS.

* An X-ray of my wrist.

* A video "voicemail" left by the caller when I don't pick up a Meet call.

* My electricity consumption as measured by the smart meter in my house.

* (the list goes on)

The fact that none of them appear in my filesystem* by default is, to me, the biggest failure of consumer IT since the Internet exists.

*Or at least in some standardized, interchangeable, mostly uniform container format.

I feel like you are trying too hard to make everything look uniform in the filesystem, at the expense of performance, expressivity, and whatnot. The whole point of having programs and applications is that they can do their stuff with arbitrary programmatic behaviors. I think the main hindrance to making all data "accessible" is the abundance of proprietary and/or underdocumented software. It would also be nice to have a pluggable file type parser/converter, which is what I assume @pyinstallwoes's sibling comment more or less describes.

Content Addressable Routing

A 'renderer' that interprets the memory location as some bitstream which turns it into an expectation of some type of shape for the bits at location.

OS Built on virtual machine over DHT-like system / Content Addressable Memory / Heck, throw in computations on there too, why compute things that have already been computed?

Basically Xanadu.

How do Content-Addressable documents work for content that is meant to be edited (ie. with changing content, versioning, …) or for content that is meant to include live data?

I like how Xanadu solved this.

Given any thing exists as a unique hash at a high level you get linked hashes but the way Xanadu approached the url structure to map to a bitstream is genius.

Besides some canonical, immutable checkpoints directly in a DHT, CRDTs could store the intermediate mutated state for you. Probably on a side "testnet".

What do you mean by the computations part?

> Heck, throw in computations on there too, why compute things that have already been computed?

Any result of a function has a return address in the globally accessible namespace distributed in a manner like DHT.

All data / code grow into a permanently distributed addressable routable form. In as much as the identity of the object having the means to be provided for by a deterministic route; which is not finite in size (allows growth for addressing for example md5/sha-256/sha-512 which would be part of the base routing protocol to access coordinates in such a system).

See: https://www.xanadu.com.au/ted/XUsurvey/xuDation.html

Buckland and Lynch are still doing the Friday seminar https://www.ischool.berkeley.edu/events/ias

Reminds me of the discussion around what is "content". Recent criticism of the word suggests that "content" is how art and creative expression gets processed into commodity, entertainment and consumption. With machine learning, "content mills" are hoping to generate endless amount of content without a (human) creator.

I also wonder what a "document" means in the context of a web browser. It includes HTML, CSS, JavaScript, as well as any media files like audio, video. The Pandora's box was to include dynamic scripting, so a document is run as an application, a kind of "hyper document" for multimedia communication and interaction.

> discussed whether, for example, sculpture, museum objects, and live animals, could be considered ‘documents.’

big "what are birds" energy https://www.youtube.com/watch?v=Kh0Y2hVe_bw

Also, every food is a salad: https://saladtheory.github.io/

@dang - this is actually pretty interesting, particularly because of when it was written (1997). I'd suggest you add the date in the tile, it's of most interest (to me at least) as a view of where we were all at in the '90s thinking about digital records.