Phase 4 cutover: swap interim raw URLs → data.quantecon.org/lectures

[mmcky]

Tracks the one-time swap of every interim data URL to the final data.quantecon.org/lectures/ form, once Phase 4 (Publishing) is live. Opened now, while the first interim link is still on a single unmerged branch, so the cutover is inventoried from the start rather than reconstructed later.

The two URL forms

Form	Example	When
Interim (in use now)	https://github.com/QuantEcon/data-lectures/raw/main/lectures/<file>	Every pilot repoint. Decoupled from hosting, survives the LFS switch, needs no DNS.
Final (this issue)	https://data.quantecon.org/lectures/<filename>	After GitHub Pages + the custom domain are live (PLAN Phase 4).

The interim→final swap is a mechanical find-replace that changes zero lecture behaviour — the bytes served are the same, only the host differs. That is the whole point of the decoupled interim form.

The gate — do not start the swap until this is true

Phase 4 must be live first: GitHub Pages deploying the lectures/ tree (lfs: true at checkout), the data.quantecon.org custom domain resolving to Pages, and CORS (access-control-allow-origin: *) verified for pyodide/JupyterLite. Phase 4 is itself blocked on a DNS decision: data.quantecon.org is currently a bare A record → 52.64.86.66 (an AWS Sydney box, HTTPS timing out) that must be confirmed decommissionable before the record is repointed. See PLAN.md Phase 4.

Swapping a lecture to data.quantecon.org before that host serves the file reintroduces exactly the 404 this project spent effort removing. Final URL second, never first.

Source of truth — do not hand-maintain a URL list

The inventory is derivable, so it is never a stale hand-kept list:

1. Greppable — every interim link matches one string. Regenerate the full inventory across the cloned consumer repos any time:

bin/foreach git grep -n 'data-lectures/raw/main/lectures/' -- '*.md' '*.ipynb' '*.py' '*.rst'

2. Manifests — each dataset's consumers: list already records repo + file for everything that points here (required on every repoint). The grep and the manifests should agree; a mismatch is the signal something was repointed without updating its manifest.

The swap recipe (per consumer file, once the gate is open)

Replace https://github.com/QuantEcon/data-lectures/raw/main/lectures/ with https://data.quantecon.org/lectures/ in each file below, one PR per consumer repo, then rebuild and confirm the page's figure hashes are unchanged (the P1 method: diff _images/<sha>.png against the live site).

Checklist (grows as pilots land)

• P1 — lingcod_msy_recovery.csv → lecture-python-intro lectures/msy_fishery.md — interim link live on main (P1 pilot: read lingcod data from data-lectures lecture-python-intro#792 merged 2026-07-17); awaiting final-URL swap
• P2 — realwage.csv / countries.csv / employ.csv → lecture-python-programming + lecture-python.myst lectures/pandas_panel.md — interim links live on both mains (P2 pilot: read pandas_panel data from data-lectures lecture-python-programming#578 + P2 pilot: read pandas_panel data from data-lectures lecture-python.myst#973 merged 2026-07-17); awaiting final-URL swap
• P3 heavy_tails / SCF — add on repoint
• P4 UNRATE — add on repoint
• remaining Phase 9 sweep — regenerate from the grep above

Part of PLAN Phase 4 / Phase 9. See QuantEcon/meta#338 (pilot) and QuantEcon/QuantEcon.manual#108 (convention).