nodespec

Status: earned its existence; its specification is drafted, and read by a reference checker. docs/adjudications/0001-mermaid-profile.md fixes the format, a profile of Mermaid flowchart syntax with declared names, and SPEC.md is draft 0, written from it. reference/nodespec/ is a checker for it (just run FILE...), and conformance/cases/ holds its conformance cases, written from SPEC.md alone, which the checker passes unchanged (just conform). existence/RESULT.md records that stock git merges a Mermaid flowchart silently wrong in constructed cases that meet a criterion fixed before they ran (existence/CRITERION.md), which with the product need meets the bar set for this kind. The first surface is named-node diagrams. A pre-registered spike, spike/mermaid/, asked whether stock git merges a Mermaid flowchart silently wrong. It has run, and both of its questions ended NO VERDICT (spike/mermaid/LOG.md §15). Real merged history put identity at risk once, and git refused that merge, so nothing went wrong silently. Structure decided 9 cases against a floor of 60. The spike is closed, and identity in real history is parked: public history rarely exposes it, as the pre-registration predicted (spike/mermaid/PRE-REGISTRATION.md §1.5).

nodespec is the node kind in kindspec — a specification and conformance suite for canvases and diagrams that must survive version control, alongside rowspec (the row kind).

Each kind is named after its unit of identity, because identity is the hard part of every one of them:

	unit	identity	status
rowspec	rows	opaque row ids	draft 0, published
blockspec	blocks	deliberately no minted ids	design pass
nodespec	nodes	named, not positional	specification draft 0, conformance cases and reference checker

The thing that makes this different from rowspec

The prior research reached a conclusion here that runs against the obvious analogy, and it should be read before anything is designed:

Absolute coordinates are not the spatial A1. The A1 defect is silent semantic corruption on clean merge; coordinates do not have it — a clean coordinate merge gives exactly both authors' moves.

What actually decides whether spatial data survives git is serialization, not addressing. See DESIGN-BRIEF.md.

Before writing anything

Read DESIGN-BRIEF.md. It carries what is already decided and measured, what is still open, and — most importantly — the one process question that must be settled first.