Canary Deployments: Why Shared State Can Turn a 1% Release Into a 100% Outage
Canary releases reduce blast radius, but shared caches and queues cause fleet-wide failures. Covers stateful failure modes and rollback patterns.
Practical guides on how the web is built, how it works, and how to ship it properly.
Three kinds of material, each written for a different moment in your work.
Implementation guides for the standards that govern professional web work. Each one covers what the standard requires, how to test whether you have met it, and the failure modes that turn up most often.
Investigations into real problems: what a dependency costs in load time, why a Lighthouse score moved, how one WCAG criterion plays out across five components.
Material to look up rather than read. Core Web Vitals thresholds, WCAG level breakdowns, HTTP caching headers, image format decision trees.
Every document in the library belongs to one of these areas.
Document setup and platform features. Doctype, charset, viewport, lang, Open Graph, canonical URL, favicons, container queries, and what every page needs before anything else loads.
Loading, rendering, and the measurements that describe them. Core Web Vitals, resource priority, caching behaviour, and what actually moves each of them.
WCAG 2.2 criteria applied to real components, and how to test them with assistive technology rather than with a scanner alone.
Headers, transport, and browser-enforced boundaries. CSP, CORS, cookie attributes, and the defaults that keep a page safe without extra tooling.
Structuring content so search and retrieval systems can extract, attribute and cite it without reconstructing it from surrounding context.
Infrastructure, deployment pipelines, containers, networking protocols, and the reliability patterns required to run web applications in production.
The documents the rest of the library assumes you have read. They are not the newest things published here. They are the ones worth reading first.
Core Web Vitals measure loading, responsiveness, and visual stability. Hit LCP ≤ 2.5s, INP ≤ 200ms, and CLS ≤ 0.1 at the 75th percentile of real users.
Use the right HTML element for the job. Landmarks like <header>, <nav>, <main>, and <footer> let assistive technologies announce structure and skip between regions.
ARIA can make custom widgets accessible, but the first rule of ARIA is don't use ARIA. Reach for a native HTML element first; add ARIA only when nothing native fits.
Cache-Control tells browsers and CDNs how long to keep a response. Use immutable + max-age=31536000 for fingerprinted assets and short or no-cache for HTML.
Choose the right script-loading attribute for every script tag: defer for app code, async for independent third-party, type=module for modern code. Bare script tags in head are always wrong.
Every <img> element must have an alt attribute. The value describes the image's purpose to screen readers, search engines, and anyone whose image fails to load.
We publish what building the web teaches us.
Web standards are public documents.
The W3C specifications, the WCAG success criteria, and the HTML living standard are available to anyone. They are written for the people implementing browsers, though, not for the people building sites on top of them, so the distance between what a specification says and what to do on Tuesday morning stays wide.
We close that distance through the work we build and maintain: products, tools, internal systems and client projects. Publishing them raises the baseline of the field we work in, and we would rather compete on execution than on keeping information scarce.
The publication exists because the work exists, not the other way around.
More about our teamDocuments come out of work we have built and shipped.
Where a document states a performance or accessibility result, it also states how that result was measured.
Every document carries a publication date. Corrections are logged rather than quietly edited away.
View corrections logWhat we will and will not publish, how sources are cited, and how commercial relationships are disclosed.
View editorial standardsOur own implementation practices, documented in the open. They are not web standards, and the distinction matters enough to state plainly.
Strict formatting rules for definitions using Q&A structures so AI engines can extract them as absolute facts.
Connecting content based on named semantic entities rather than just keywords, building a robust Knowledge Graph.
Building persistent deep-links onto every paragraph to encourage direct AI source citations.
Designing semantic HTML tables specifically so they can be perfectly parsed and reasoned over by LLMs.
Enforcing time-stamped, machine-readable transcripts alongside all video/audio content.
Strict engineering rules for reserving exact DOM space for images to guarantee a 0.0 CLS.
Infrastructure, deployment, and monitoring practices. This content lives separate from web standards, focusing on how we build and ship reliably.
Blue-Green deployments switch traffic instantly between two identical environments. Canary releases shift traffic gradually to a subset of users to limit the blast radius of a bad deployment.
Understand RPO (Recovery Point Objective) and RTO (Recovery Time Objective). Direct calculation formulas, real-world failure scenarios, and cost-tier trade-offs.
Automated probes used by load balancers to determine if a backend server is capable of handling traffic, preventing requests from being routed to failed instances.
How to implement exponential backoff and jitter in retry logic — the technique that prevents cascading failures when a downstream service is struggling under load.
Explicit, step-by-step technical manuals designed to guide developers through mitigating specific alerts and routine operational tasks.
How TCP keepalive probes work at the OS level, why cloud firewalls silently drop idle connections before the default 2-hour timeout, and how to configure tcp_keepalive_time to prevent application deadlocks.
The newest implementation guides added to the library. Each one explains what the standard requires, how to implement it, and how to verify it in practice.
If a domain is genuinely for sale, say so in DNS. A TXT record at _for-sale.example.com advertises it to brokers and availability services without taking the site down or parking it.
A dead header that roughly a third of major sites still send. No shipping browser reads it, it was never standardised, and the values people copy-paste were the dangerous ones. Stop sending it and rely on CSP.
If you publish a feed, advertise a WebSub hub with rel="hub" and rel="self" so subscribers are pushed each update instead of polling for it. Cuts latency to seconds and removes most of the crawl traffic your feed attracts.
In-depth architectural analyses, browser engine internals, and production runbooks for modern web infrastructure.
Canary releases reduce blast radius, but shared caches and queues cause fleet-wide failures. Covers stateful failure modes and rollback patterns.
How Claude Search works: Brave Search indexing, BM25 term weighting, dense vector embeddings, and structuring self-contextualizing passages.
Deploy RFC 8615 /.well-known/ endpoints without silent CDN failures. Covers RFC 9116 security.txt, A2A agent-card.json, and MCP routing.
Why JavaScript focus traps break under CSS reading-flow, scroll containers, and Shadow DOM, and how native dialog and inert provide the 2026 fix.
A 403 on /robots.txt grants AI crawlers full access. Covers a two-minute diagnostic, 12 silent production failures to fix, and bot behavior auditing.
We build to the same criteria published here, with Core Web Vitals judged against field data, WCAG 2.2 AA verified with assistive technology, and architecture documented as it's built.
{
"webSpecification": {
"coreWebVitals": "field data",
"accessibility": "WCAG 2.2 AA",
"architecture": "documented as built"
}
}