Data coverage

What we poll, how often, and what we deliberately don't promise. Honest staleness beats fake real-time - here's where every fact on a /d/<name> page comes from.

The universe in three tiers

We index roughly 49 million gTLD domain names. We do not poll them all every day - nobody does, and anyone claiming to is reading a stale cache. Instead we split the universe into three tiers and assign each tier a polling cadence we can actually sustain.

Hot tracked

~100K-500K domains

Domains in active drop cycle (expiring, grace, pending-delete, just-dropped), domains visited recently, and domains flagged by threat feeds. Refresh cadence: WHOIS, Wayback and reputation are fetched on demand when a page is opened; the drop cycle is ingested daily after 06:00 UTC.

Cold tracked

~1-5M domains

Domains we've polled at least once but that aren't in active cycle. We do not refresh them on a schedule - we refresh them when you visit /d/<name>. The badge at the top of every domain page tells you exactly when the last fetch happened. If it's stale, just opening the page enqueues a fresh poll.

Zone index

~49M names

The bulk of the gTLD universe: names we know exist (from ICANN zone files) and the date we first saw them. We keep this index on disk, not in the database - a tracked record is created only when something happens to the name: it drops, enters an expiry cycle, or gets enriched. Until then /d/<name> reports it as not tracked.

Sources and cadence

Every signal you see on a domain page comes from one of these. We log every observed change to a per-field append-only history table - if we poll a source and it differs from the last value, that becomes a timestamped row you can read on /d/<name>.

Source Cadence (hot) Cadence (cold) Cost
ICANN CZDS - daily zone-file diff. Detects drops + new registrations across 14 gTLDs. Daily Daily Free
WHOIS / RDAP - registry-direct lookup for status, registrar, dates, nameservers, EPP codes. 1-7 days On visit Free
Wayback Machine CDX - snapshot counts, first / last capture, archived <title>. 30 days On visit Free (rate-limited)
Spamhaus DBL - live phishing / malware / botnet listings, with state-change timeline. Daily On visit Free
DNS resolver - A / MX records. Logged when changed. Daily On visit Free
crt.sh CT logs - active certificate count, unique subdomains, first-seen year. Monthly On visit Free
Wikipedia external links - mention count from API. Monthly On visit Free
Hacker News (Algolia) - story / comment mention count. Monthly On visit Free
GitHub Code Search - mention count across public repos. Monthly On visit Free (token-pooled)
Reddit - mention count across all subreddits. Monthly On visit Free

What we don't promise

Why this tiering exists

Polling 49 million domains daily for every signal would cost real money and break every source's rate limit within minutes. Polling 49 million domains monthly still produces a 6-week-stale snapshot for any domain you care about. The honest answer is to spend our polling budget where it matters - on domains people are actually looking at right now - and to surface the freshness state plainly so you can tell stale data apart from fresh.

If a domain you care about isn't getting refreshed often enough, opening its /d/<name> page enqueues a fresh fetch on the on-demand worker pool. The page itself updates as data lands via Turbo Frames.

← Back to the homepage