Split into two stages to allow IPLD to develop further:
- Now: JSON-LD file-based model. JSON-LD is the future of RDF and it's already popular. The simplest, most future-friendly and semantically-idiomatic model to get the Underlay started now is to publish JSON-LD files that:
- are flattened relative to one top-level context,
- address every node with either absolute or fragment-relative URIs (no blank or path-relative nodes),
- link to other nodes with the IPFS URI scheme
dweb:/ipfs/Qm..., using fragments for nodes within documents:dweb:/ipfs/Qm...#node-id - use the PROV vocabulary with named graphs to capture "first-class" provenance as JSON-LD itself
- are persisted to IPFS with the
filesAPI, and