Glossary

The lang attribute states what language a document is written in. Who uses it, how to mark passages in other languages, and what a wrong value does.

Alt text conveys an image's purpose to anyone who cannot see it. When it should be empty, what to leave out, and why it is not a caption.

A favicon identifies a site in tabs, bookmarks and search results. Which files browsers look for, what sizes matter, and why caching is stubborn.

A CDN caches content at edge locations near visitors. What it improves, what it cannot improve, and what changes when one sits in front of a site.

Core Web Vitals are LCP, CLS and INP, measured on real visits. What each covers, what the thresholds are, and how field and lab data differ.

LCP measures when the largest visible element finishes rendering. What counts as an element, the four phases of the delay, and what usually dominates.

CLS measures content moving without user action. How the score is calculated, what causes shifts, and why reserved space is the fix.

INP measures the latency of interactions across a whole visit. What the three phases are, why it replaced FID, and what usually causes a bad score.

FCP marks the first moment any content is painted. How it differs from LCP, what blocks it, and why it matters even though it is not a Core Web Vital.

TTFB measures from navigation to the first byte of the response. What the phases are, what it does and does not include, and what a high value means.

Compression shrinks text responses in transit. How negotiation works, how gzip and Brotli compare, and why images should not be compressed again.

Cache-Control states who may store a response and for how long. What each directive means, and why immutable assets and HTML need opposite policies.

An ETag identifies a specific version of a resource so a cache can revalidate cheaply. Strong and weak tags, and why they break behind load balancers.

rel=preload tells the browser to fetch a resource it would otherwise discover late. When it helps, what the as attribute does, and how it is misused.

rel=preconnect performs DNS, TCP and TLS setup for an origin ahead of time. When it saves real time, and why using it widely is counterproductive.

rel=dns-prefetch resolves a third-party hostname before it is needed. What it saves, how it compares with preconnect, and when it is worth declaring.

A render-blocking resource must be loaded before the browser paints anything. Which resources block, why CSS does, and what defer and async change.

A large DOM makes every style recalculation and layout pass expensive. What the practical limits are, what causes bloat, and what it costs.

Page weight is the sum of everything a page fetches. What typically dominates it, why compressed and uncompressed sizes differ, and what it costs.

Lazy loading defers offscreen resources until they are needed. How the loading attribute works, and why lazy-loading the hero image backfires.

Above the fold is the part of a page visible without scrolling. Why it has no fixed size, and why it decides which resources are critical.

A third-party script runs in your page with full access to it. What it costs in performance and risk, and what containment options exist.

What the fetchpriority attribute changes, why it is not a substitute for removing a lazy attribute, and where raising priority costs something.

The sequence a browser works through before it can paint, which resources block it, and why late discovery costs more than file size.

What total blocking time measures, why only the part of a long task beyond fifty milliseconds counts, and how it relates to interaction responsiveness.

How splitting a bundle changes what a visitor downloads and parses, why parse cost matters more than transfer, and where splitting goes too far.

What minification strips, why its gain is smaller than compression's, and why an unminified bundle in production usually means a build misconfiguration.

The choice between invisible and swapped text while a web font downloads, what each value does, and why the fallback metrics matter as much.

Duplicate content is the same or near-identical content reachable at several URLs. Why there is no penalty, and what consolidation actually costs.

Crawl budget is the number of URLs a search engine will fetch from a site. What limits it, which sites it affects, and what actually wastes it.

Content built by JavaScript is invisible until a client executes it. Which consumers render and which do not, and what that means for a page.

A canonical tag names the preferred address for content reachable at several URLs. Why it is a hint rather than an instruction, and when a redirect is the better answer.

What an hreflang annotation declares, why every reference must be reciprocal, and the three mistakes that make a whole cluster be ignored.

What a soft 404 is, why returning 200 for a missing page wastes crawl budget, and the three site patterns that produce them at scale.

What the X-Robots-Tag header does that a meta tag cannot, why it is the only option for non-HTML files, and how it is accidentally applied site-wide.

What makes a page an orphan, why a sitemap entry is not a substitute for a link, and why a crawl-based finding is sometimes a false positive.

What click depth measures, why the three-click guideline is a heuristic rather than a rule, and what actually makes a deep page hard to find.

What a sitemap index is, the limits that force one, and the reporting trade-off between a single sitemap and several.

What an HTML sitemap is, how it differs from the XML one, when a site is large enough to need one, and why most sites are not.

A breadcrumb trail shows a page's position in a hierarchy. What the markup adds, how search results use it, and what makes a trail honest.