CMS snapshot distribution plan

Companion to MIGRATION.md and INVENTORY.md. Covers moving the 22 extracted CMS conference snapshots into their target ESL repos. All 22 are extracted, audited, and content-complete (0 download failures, 0 missing local assets); see the CMS-extraction audit summary at the bottom.

Status: new repos created; new-repo push scripted & dry-run-verified

UPDATE 2026-07-06: the three new repos exist and are empty (esl/code-beam-lite, esl/code-mesh, esl/code-elixir-ldn). Their 13 editions are scripted in migration-data/distribute_new_repos.sh (dry-run verified: correct repo/dir mapping, all snapshot sources present). RUN IT IN AN AUTHENTICATED TERMINAL: bash migration-data/distribute_new_repos.sh (or DRY_RUN=1 bash ... to preview). The 9 editions for the existing repos (code-beam-america, code-beam-stockholm) are scripted in migration-data/distribute_existing_repos.sh (dry-run verified; pushes a BRANCH and prints the PR URL — never the default branch). Run it in an authenticated terminal after the new repos are verified. CNAME/DNS still pending sysadmin.

Status: PREPARED, blocked on GitHub write access

This automation session has no GitHub write access: no gh CLI, and the macOS keychain is unreadable headlessly (git push → “could not read Username … Device not configured”). So it cannot create the new repos or push branches. Everything below is prepared so an authenticated user (or session) can execute it mechanically. Per MIGRATION.md: PRs only, never push to a brand-repo default branch.

Where the snapshots are

Session scratchpad: <scratchpad>/snapshots/<slug>/ (temporary, git-ignored, regenerable). Each is a self-contained static folder: index.html, speaker/<name>/index.html, media/<slides>/index.html, assets/..., REPORT.txt. Regenerate at distribution time with the current migration-data/extract_cms.py (now includes nav-hardening — see below) on a stable connection:

python3 migration-data/extract_cms.py <all 22 slugs> --out <dir>

The tool is idempotent (skips already-downloaded assets), so an interrupted run resumes safely by re-invoking with the same args.

Convention finding (how existing archives are built)

Existing archives — code-beam-europe/archives/berlin_2023, code-beam-america/archives/CBA_2024 — are full-site wget-style mirrors: the entire site chrome (participants/, talks/, keynotes/, nav pages) duplicated per edition, with root-relative links and the repo’s own CNAME.

These snapshots are leaner and self-contained instead: the conference page plus its speaker and slide subpages, assets localized to relative assets/ paths (works offline), and cross-site hub nav links (/about-us/, /sponsorship/, etc.) rewritten to absolute https://codesync.global/... so they still resolve when served from a different domain. Internal folder names follow the CMS URL structure (speaker/, media/) rather than the Jekyll site’s (participants/, talks/).

DECISION NEEDED: accept the leaner self-contained format (recommended — renders anywhere, no dependence on a live CMS), or re-capture as full-site mirrors to match the existing archives exactly.

Target mapping

Existing repos — drop each snapshot into the repo’s archive dir (note: code-beam-america uses archives/, code-beam-stockholm uses archive/ singular). Proposed edition dir names follow each repo’s existing scheme; confirm naming before pushing.

Snapshot slug	Real edition	Target repo	Proposed archive dir
code-beam-sf-2018	Code BEAM SF 2018	code-beam-america	archives/sf_2018
code-beam-sf-2019	Code BEAM SF 2019	code-beam-america	archives/sf_2019
code-beam-sf	Code BEAM SF 2020	code-beam-america	archives/sf_2020
code-beam-v-america-2021	Code BEAM V America 2021 (Mar)	code-beam-america	archives/v_america_mar_2021
code-beam-sf-2021	Code BEAM America 2021 (Nov)	code-beam-america	archives/america_nov_2021
code-beam-sto-2019	Code BEAM Europe 2019 (Stockholm)	code-beam-europe	archives/stockholm_2019
code-beam-sto	Code BEAM Europe 2020 (virtual)	code-beam-europe	archives/virtual_2020
code-beam-sto-2021	Code BEAM Europe 2021 (virtual)	code-beam-europe	archives/virtual_2021
code-beam-sto-2022	Code BEAM Europe 2022 (Stockholm)	code-beam-europe	archives/stockholm_2022

NOTE: these four are the FLAGSHIP Code BEAM Europe (Stockholm/virtual era), NOT Code BEAM Lite Stockholm. codebeamstockholm.com hosts the separate Code BEAM Lite Stockholm brand (2023+) and must not receive these editions.

New repos — create first (see below), then add editions:

Snapshot slug	Real edition	Target repo (NEW)	Proposed archive dir
code-beam-lite-berlin-2018	Lite Berlin 2018	code-beam-lite	archives/berlin_2018
cbl-amsterdam-2018	Lite Amsterdam 2018	code-beam-lite	archives/amsterdam_2018
cbl-munich-2018	Lite Munich 2018	code-beam-lite	archives/munich_2018
code-beam-lite-italy	Lite Italy 2019	code-beam-lite	archives/italy_2019
code-beam-lite-budapest	Lite Budapest 2019	code-beam-lite	archives/budapest_2019
code-beam-lite-berlin-2019	Lite Berlin 2019	code-beam-lite	archives/berlin_2019
code-beam-lite-india	Lite India 2019 (partial)	code-beam-lite	archives/india_2019
code-beam-lite-amsterdam	Lite Amsterdam 2019	code-beam-lite	archives/amsterdam_2019
code-beam-lite-virtual	Lite Virtual 2020	code-beam-lite	archives/virtual_2020
code-mesh-2018	Code Mesh LDN 2018	code-mesh	archives/ldn_2018
code-mesh-ldn-2019	Code Mesh LDN 2019	code-mesh	archives/ldn_2019
code-mesh-ldn	Code Mesh V 2020	code-mesh	archives/ldn_2020
code-elixir-ldn-2019	Code Elixir LDN 2019	code-elixir-ldn	archives/ldn_2019

New repos to create (needs auth + decisions)

esl/code-beam-lite — orphan Lite cities. No single live domain. CNAME DECISION: new subdomain (e.g. lite-archive.codesync.global), a path under an existing domain, or github.io default.
esl/code-mesh — brand retired. codemesh.io still alive (2016/2017); CNAME could be codemesh.io if that domain is under ESL control and repointed.
esl/code-elixir-ldn — brand retired; elixir.london DNS is dead. CNAME DECISION needed (new subdomain or github.io).

Model each new repo on an existing one’s serving setup (static files + Pages). Since snapshots are self-contained static HTML, the repo needs no Jekyll build — just the archives/<edition>/ folders + CNAME + Pages enabled.

Per-edition push steps (for the authenticated operator)

For existing repos (example, code-beam-america):

git clone https://github.com/esl/code-beam-america.git
cd code-beam-america
git checkout -b archive-cms-editions
cp -R <snapshots>/code-beam-sf-2018   archives/sf_2018
cp -R <snapshots>/code-beam-sf-2019   archives/sf_2019
# ... etc for this repo's rows
git add archives && git commit -m "Add CMS-era archive editions"
git push -u origin archive-cms-editions
# open PR against default branch

For new repos: create the repo, add a CNAME + README, then the same cp -R of each edition into archives/, commit, push, enable Pages.

Remaining manual cleanup before/after pushing

code-mesh-ldn: one inline “buy your ticket on Eventbrite” text link with a dead 2020 URL in prose (not caught by section-level stripping). Snip by hand.
Verify a couple of pushed archives render on Pages (Pages path-serving vs the relative asset paths — should work since assets are relative, but confirm).

Hub-side follow-up (separate from brand repos)

After archives are live, add/point hub /conferences/ entries for these editions (most have no _conferences/*.md yet — INVENTORY.md “Hub repo data fixes” section). Extracted editions link out to their new archive URLs, matching how ElixirConf EU / Lambda Days rows already link out.

CMS-extraction audit summary (2026-07-06)

22/22 conferences extracted. 0 download failures, 0 failed subpages, 0 missing local assets (every speaker photo, bio, schedule, slide PDF, css/js/font, and the ckeditor syntax-highlighter resolves). Strip rules verified: only registration/ticket/newsletter/cookie sections and Eventbrite/analytics scripts removed. code-beam-lite-india correctly has no schedule (its CMS page was never built out). Visual spot-checks (Code BEAM Lite Italy, Code Mesh 2018) render offline faithfully with schedules, speaker photos, talk-page video embeds, and working slide downloads.