ADR 0001 — ROS Markdown remains canonical

Status: Accepted (constraint, §1.1)

Context

The corpus is 837 Markdown files across two populations (B.0). ROS governs it through ros.json, registries and the work protocol.

Decision

ROS Markdown is the sole canonical research record. The publisher is read-only with respect to it. Typed runtime models, derived indexes and generated HTML are projections, never replacements.

Evidence

OBSERVED: every relationship, status and identity in the corpus originates in Markdown front matter or body text. No other store exists. DOCUMENTED ROS RULE: ros validate and the registries already own artifact governance (D.4).

Alternatives

Convert to JSON or a database — rejected by §1.1 and unjustified: 756 documents build in 5.6 s, so no performance argument exists.

Consequences

The parser must absorb all corpus heterogeneity (D.1). This is the cost of the constraint and is accepted.

Reversibility

Reversing would mean a migration of canonical research — expensive and explicitly out of scope.