About this archive

What this is

The History of Hosting is a chronicle of the web-hosting industry drawn from WebHostingTalk (WHT), the forum that has served as the trade press, complaint department, and wedding registry of the hosting business since 1998. The site you are reading is generated automatically from a live crawl of that forum, then written up — partly by software, partly by an LLM, always grounded in the posts themselves.

Why "popular threads"

WHT holds roughly 1.45 million threads and 8.6 million posts. Crawling and reading every one would take days and cost more in compute and attention than the marginal history justifies. Instead we target the high-signal core: the highest-reply and highest-view threads across all 157 forums — the acquisitions, outages, scandals, launches, and pricing revolts that the community itself elevated to the top. This yields 23,772 candidate threads; the archive currently holds 1,372 of them (259,621 posts), growing as the crawl proceeds.

How it was built

Discovery. vBulletin exposes sort=replycount and sort=views on its forum listings. We walk every forum's top threads down to a threshold of 50 replies (and the top 10 view pages), deduplicating by thread id.
Crawl. Each thread is fetched through Cloudflare with a challenge-solving client, paginated in full (capped at 30 pages per thread for breadth), and parsed into structured posts — author, timestamp, body text with quoted replies stripped.
Analysis. A Python pass extracts, per year: company and technology mentions (curated lexicons), advertised plan prices and specs (regex), and thread counts. These feed the charts and the per-year "Mentioned" tags.
Narrative. For each decade, a source pack of flagship threads, opening-post excerpts, and quantitative facts is assembled from the live database and handed to a language model under a strict editorial style brief. The brief forbids the usual AI filler ("ever-evolving", "testament to", "delve", "realm") and requires every claim to be anchored to a specific thread, price, or year from the source pack. Nothing is fabricated; if the crawl didn't capture it, it isn't said.
Site. A Jinja-free Python generator renders the HTML and hand-tuned inline SVG charts. No JavaScript frameworks. Quotes link to local archive pages, which in turn link back to the live WHT thread.

On attribution and honesty

Every pull-quote names its author (where recorded) and links to the thread it came from. The decade essays are regenerated hourly as the crawl accrues, so they reflect the current state of the archive rather than a frozen moment. The 2020s are thinner than the 2000s because the crawl prioritised the highest-reply threads first; coverage fills in over time and the prose says so when the data is thin.

Built on an exe.dev VM. The crawl is polite, resumable, and completed without errors. This page, like the rest, is regenerated automatically.