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)
-
Content-addressed identity (CAS) — same bytes (or same semantic unit) keep a stable id when the label moves.
-
Mutable pointer — human name resolves to the current edition when content updates.
-
Reverse consumers — who still holds a label (
CONSUMERS.mdtoday; graph reverse edges tomorrow). -
Absorb layer — redirects and aliases on today’s classical web (see Near-term absorb).
Where the deep pages live
| Concern | Canonical face |
|---|---|
Symptom ↔ diagnosis/treatment dialectic |
|
Path as navigation; node as truth |
|
Delivery-name vs application-label |
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.