Identity stack

Internet Reliability’s treatment stack is the same claim HCI Nerdz and connectome-fs already teach. DevCentr’s job here is orientation at systems altitude, not a second full treatise.

flowchart TB
  Consumer[Consumer · docs · agent]
  Hash["Content hash / node id\n(immutable identity)"]
  Pointer["Mutable human pointer\n(label → current hash)"]
  Absorb["Absorb layer\n(redirects · aliases)"]
  Consumers["Reverse consumers\n(CONSUMERS.md / graph)"]

  Consumer -->|"prefer"| Hash
  Consumer -->|"or resolve"| Pointer
  Pointer --> Hash
  Absorb -.->|buys time on IP web| Consumer
  Consumers -->|"when labels must move"| Consumer

Four layers (one stack)

  1. Content-addressed identity (CAS) — same bytes (or same semantic unit) keep a stable id when the label moves.

  2. Mutable pointer — human name resolves to the current edition when content updates.

  3. Reverse consumers — who still holds a label (CONSUMERS.md today; graph reverse edges tomorrow).

  4. Absorb layer — redirects and aliases on today’s classical web (see Near-term absorb).

Diagram

Where the deep pages live

Concern Canonical face

Symptom ↔ diagnosis/treatment dialectic

HCI — Labels versus wires

Path as navigation; node as truth

connectome-fs

Delivery-name vs application-label

NDN and friends

Math sketch (same as HCI): if bytes are unchanged under rename, \(Hash(Bytes_{old}) = Hash(Bytes_{new})\). Edits yield a new hash; a mutable pointer must advance, or consumers freeze on an old edition.