RE Redirector

Over twenty years I have collected domains: the bare thingelstad.com, the micro.thingelstad.com the blog used for a while, weeklything.com, jamiethingelstad.com, and thinglestad.com with its letters swapped, the way people keep typing my name. Each needs to answer and send the visitor somewhere useful. For years that was a scatter of registrar redirects. In September 2026 I began moving them into one CloudFront distribution with a small function in front of it, and by late October every one of them was answered in one place.

It started as plumbing: a name in, a 301 out. A redirect keeps the path, so a deep link to an old host still reaches the same page on the new one, and it adds a ref naming the host it came from, so my analytics show which old names people still use.

On October 31, 2026 it started making decisions. The blog has moved from .Text to WordPress, Squarespace, Jekyll and Micro.blog, and links out on the web still point at addresses that stopped existing years ago. Sites that linked to them will never update, and search engines kept finding 404s. Now the function looks each old blog path up in a table of old addresses matched to their posts and sends the visitor straight to the post. The table started with 905 matches, made only where the old slug and date clearly point at one post. Each Saturday the paths still landing on a 404 are ranked by how many real visitors follow them, so I add matches where people actually arrive.

About a quarter of the requests were never people: scanners probing for wp-login.php, xmlrpc.php and .env files. They get their own answer, an HTTP 418 “I’m a teapot” page explaining that this house hasn’t run WordPress since April 20, 2016, with a haiku in the source for anyone who looks. Every request now records what the function decided (a mapped post, a teapot, a plain redirect, or a fixed page) in the access log, so I know concretely what it is doing.