# Specification
1. The library measures only the current revision of every canonical Kweb
node discoverable beneath the root's node store.
2. Full-node totals contain the short name, short description, and long
description.
3. Long-description totals contain only long descriptions.
4. Character counts count Unicode scalar values. Word counts use Unicode
whitespace boundaries as implemented by `str::split_whitespace`, with each
node text field counted independently.
5. Estimated tokens equal the character count divided by four, rounded up.
6. History, provenance, connections, ownership, objects, and all non-node data
are excluded.
7. Results are returned as typed values. Human-readable rendering is a
separate operation.
8. Measurement rejects a missing root rather than creating a new Kweb.
9. The library owns no CLI, credential vault, process lock, or server policy.