Semantic Paint will be a system for annotating text that does not rely on filenames or URL's or any other naming system. Instead it uses fuzzy hashes to identify where in the text to attach the annotation. If you peer your node with other people, you'll see their annotations too.
It's like the web, except you don't need permission to create links between data, and whether you see other people's links/data depends on whether you explicitly subscribe to their "brushstrokes" in that "color".
"Colors" might include:
- this-javascript-is-malicious-and-here-is-a-patch-that-neuters-it
- this-is-a-claim-and-here-is-evidence-for-it
- the-food-described-by-this-menu-entry-causes-allergic-reactions-for-people-with-this-kind-of-allergy
Ideally, users (or aggregations thereof) will emerge as authorities on these topics and we can figure out how to pay them for being useful curators.
Since the references (this and here in the above example) are anchored to other pieces of text by CTPHs and not by name, you can find text in the world and point your camera at it and now you have links to follow.
There's a mapping between CTPH identified text and cryptographic-hash identified text, so we can use a DHT to move content around, I'll probably have plugins for different network constraints (i.e. Bittorrent, IPFS, Freenet, SSB, etc).
Since I'm solving the hard problem of naming things by just not naming things, I don't need an always-available DNS server (for instance), so I might as well see what else I can do without. I'm shooting for partition tolerant and latency tolerant, to run over secure sucttlebutt protocol (which presumes that the user is moving around gossips over ad-hoc networks), so it can be useful in a world where the ISP's are unusable for some reason.