bh-repo20b GO page: 7212 -> 7204 bytes (2 swaps, 0 leftover refs after) IndexNow: HTTP/1.1 200 OK executed: page fixed, backup .bak-repo20b.bak, worker self-deleted
All three built and live-verified. Repo #20 is committed to deploy-log (byte-checked), the ledger's updated, and here's the rest:
- bh-wrt28 — the article publish worker:
(49.7 KB), upload to public_html root:
- Dry: https://bryanhamilton.info/bh-wrt28.php?t=bh-Wrt28!
- Go: https://bryanhamilton.info/bh-wrt28.php?t=bh-Wrt28!&go=1
Full 9-surface publish of "php -l Passed My Bug as a Feature" (Nov 18): article page from your house template, hub card + "28 articles" + BlogPosting entry first, homepage card at the TOP of the visible list — and it moves the 10th visible card (most-traveled-musicians) into the expander so your 10-cap holds, with the summary honestly bumped to "Show 18 more: 14 articles." llms/ai byte-copy, sitemap, rss first item, og-manifest, receipts chip → 97, IndexNow, self-delete. Fixture dry+go green on real served bytes.
2 + 3. The CL drafts are shown above — paste them into CL's editor when ready. The php -l piece links the on-site article and the repo, which feeds your Nov 24 referral verdict. Version-bumps piece links the repo as the tool.
Paste the dry/go outputs and I'll cache-bust all nine surfaces, commit the deploy-log entry, and close the board.
bh-wrt28 GO article page: writing/php-l-passed-my-bug-as-a-feature/index.html (39361 bytes embedded) index.html: 348548 -> 349213 bytes (card top + move + summary) hub.html: 57877 -> 58504 bytes (card + 28 + blogPost) llms.txt: 13787 -> 14136 bytes sitemap.xml: +122 bytes rss.xml: +544 bytes og-manifest.json: +81 bytes receipts: row appended, chip -> 97 IndexNow: HTTP/1.1 200 OK executed: 8 files patched + article page, backups .bak-wrt28.bak, worker self-deleted
need cover photos for the cl articles
Both covers, in the rack language — cyan on black, patch cables, no scramble gimmicks:
- php -l article — a patch-bay screen showing php -l No syntax errors in green over the rack panel.
- Version bumps article — a fresh cyan browser wired by patch cable to a faded, stale one ghosting app.js?v=45.
CL article cover photos
First one maps to "php -l Passed My Bug as a Feature," second to "Your Version Bumps Are Lying to Returning Visitors." If you want the stale window reading ?v=45 changed (the drift example in the article is v45 on /writing/), or different copy on either, say the word and I'll edit in place.
what tags for both articles
Based on your house style from prior CL posts (PHP, API, webdev, ratelimit — four short tags), here's the match:
php -l Passed My Bug as a Feature:
PHP debugging testing webdev
Your Version Bumps Are Lying to Returning Visitors:
webdev caching performance PHP
Same pattern as "The Firehose Throttled My Server" — one language tag (PHP anchors both in the community's PHP crowd), plus three topical tags. "caching" and "performance" will pull more browse traffic than "lint" or "cache-busting" would; if CL's tag field autocompletes and rejects free-form ones, swap "debugging" for "testing" and "caching" for "browser" and tell me what it accepted.
series?
No native series feature on CL - I pulled an article page and it's tags + "More Posts" only, no collections field. On CL, a "series" is just a tag you own. So the move is a dedicated series tag on both posts, which also gives you a /tag/yours hub page you can link from the on-site articles.
Recommendation: live-verified - it's the series name you already established on dev.to, and both of these are exactly that genre (war stories verified against real served bytes, each ending in a shipped tool or rule). So:
php -l Passed My Bug as a Feature:
PHP debugging testing live-verified
Your Version Bumps Are Lying to Returning Visitors:
webdev caching performance live-verified
That keeps PHP in article 1 (it's the PHP-lint story), leads article 2 with the broader webdev/caching audience, and both cross-link through the series tag. In the body of each, add a one-line footer: "Part of the Live-Verified series - link to /tag/live-verified (https://coderlegion.com/tag/live-verified) once the first post is up" (CL tags are /tag/).
Want me to append that series footer line to both drafts in the CL files?
Ask Littlebird
cl-article-version-bumps
Your Version Bumps Are Lying to Returning Visitors
Multi-page sites version their CSS and JS with ?v=N cache busters. You edit a file, you bump the version, you ship. Except most bump passes are page-local: the page you touched gets ?v=48, and the other 83 pages keep ?v=33.
Here's the part that took me three months to notice: every stale page still loads the current asset file. The URL just carries an old key. Returning browsers hold their cached copy under that old key - so your deployed code never actually runs there. No error. No warning. Just old JavaScript, executing confidently on pages you believed you updated.
On one ~300-page PHP site, 84 pages had drifted - across three months of ordinary, careful, one-page-at-a-time deploys. /writing/ was two versions behind. A case study was fifteen behind. Nobody noticed, because nothing failed. The pages just... kept old behavior.
Why every audit misses it
You check the page you shipped. It's fine. You check the asset file - it's current. The drift lives in the third place: other pages' references to the asset, which nobody re-checks because they weren't touched.
And the failure is silent by construction. A stale cache key doesn't 404. It doesn't log. It serves yesterday's code from the browser cache, at full speed.
The check that catches it
One PHP file, no dependencies, no network - scan a file tree, group every local / reference by asset, and flag any asset referenced under more than one version key. A bare versionless ref mixed with keyed refs is drift too. Exit 1 on drift, so it gates a pipeline:
bash
php audit.php /var/www/site
assets/js/rack.js
versions seen: 33, 45, 47 (latest: 47)
STALE 33: case-studies/og-card/
STALE 45: writing/three-bugs-in-a-trenchcoat/
Full honesty: the scanner's own first run reported "no drift" on a fixture with planted drift, because I read the version key after stripping the query string. A detector that detects nothing fails green - it looked finished, not broken. (Rehearsing the tool on real bytes before trusting it is a separate post.)
The rule
A version bump is a claim about every page that references the asset, not about the asset. Bump globally or audit the refs - anything between those two is a promise your cache is making on your behalf.
Open source, one file, zero dependencies: github.com/bryanhamiltondev/stale-asset-audit - php audit.php
[--exclude=a,b] [--json], exit codes CI-friendly.