RFC: pick the address a content API caller uses to name a node #163

Problem

Before the content API can target part of a document, one word has to name the target. Today headings and tables carry a toc-id attribute, and no other node has an identity.

Whatever is chosen becomes the address used in deep links, so it has to survive edits.

What to decide

Which address a caller uses: the existing toc-id, a root index, or a new node id. Each has a different lifetime under editing.

Also what a "section" means on the server. The document schema is flat, so a section is already a run of root siblings. The open part is packaging: promote the webapp's section helper into a shared package, or write it again on the server.

Acceptance

• One address is named, with its limits written down. toc-id covers headings and tables only.
• The section helper is either promoted or duplicated by decision, not by accident.

Blocked by

Only worth answering if node targeting is approved. See that RFC.