If you've ever followed a population citation on Wikipedia and landed on nothing, there's a decent chance it pointed here.
From 2001 to 2013, this domain published the World Gazetteer: Stefan Helders' database of city and country populations, covering about 148,000 places. Spanish and Portuguese Wikipedia leaned on it especially heavily. Thousands of those citations are still live, still pointing at URLs that stopped working thirteen years ago.
What made it good
Most population sources give you one number per place and leave you to work out what it means. The World Gazetteer gave you three, and labelled them.
It separated the city proper from the urban agglomeration from the administrative division. London was 7.4 million as a place and 12.6 million as an agglomeration, and the database said which was which. That sounds like a small thing. It's the single biggest source of confusion in population data, and twenty years later most sites still get it wrong by not saying.
It also had a downloadable file. A whole database, free, no registration. That generosity is why it ended up cited everywhere.
Then it stopped
The site went offline around 2013. Helders stopped maintaining it, the hosting lapsed, and at some point the domain fell into the hands of a parking service that ran ads on it.
Normally this is where the Internet Archive saves the day. Not here.
The World Gazetteer had been fighting scrapers for years, and its defence was to block automated readers with a terse message in Spanish: no se permite el uso de un Offline-Reader. Consult the downloadable file instead.
The Wayback Machine is an automated reader. So what archive.org holds, capture after capture, year after year, is that block message. The data pages were never preserved. A reference work cited across Wikipedia in a dozen languages had effectively deleted itself.
The file that survived
Except for one thing. That block message told crawlers to go and download the data file instead, and on 29 December 2005, something did.
A single capture of dataen.zip, 2.3 MB, sitting in the archive under a URL nobody had reason to look at. The Wayback Machine's own index records it: timestamp 20051229193956, content type application/zip, 2,348,697 bytes, and every later capture of the same URL is a redirect from the domain-parking years. Inside: one tab-separated file, 148,368 rows, name and population and type and country and region and coordinates, dated 18 June 2005.
We checked it against known figures. England 50,233,280 against a published mid-2005 estimate of about 50.4 million. Tokyo's agglomeration at 36.5 million, matching the UN. China at 1,303,458,527. It's accurate, and it's the only public copy we've been able to find.
Why it's back here
We bought this domain to build a population site, and then discovered what it used to be. Restoring the archive was the obvious thing to do.
So the 2005 edition is now online at the World Gazetteer archive: 237 countries, each with its cities, agglomerations and administrative divisions listed separately, the way Helders organised them. The complete dataset is a free CSV download, credited to him, as it should be.
The old citation URLs work again too. A link from Spanish Wikipedia carrying wg.php?geo=-45 now lands on Canada's archive page instead of an error.
One thing we won't do
Those pages serve 2005 figures, clearly labelled, and nothing else. It would be easy to point thousands of old academic citations at today's numbers and take the traffic. It would also mean Wikipedia articles citing us as the source for figures our pages don't contain.
So the archive stays an archive. If you want current numbers, every country page links across to them, and that's a click the reader chooses to make.
Stefan Helders built something careful and gave it away. The least we can do is not misrepresent it.