Redirect and HTTP header checker

If the old URL is not coming back, 301. A 302 tells Google the change is provisional, so it keeps treating the old URL as the original and takes far longer to pass the signal to the new one. Almost every 302 we see in migrations is a misconfigured 301: the default on many platforms is 302 and nobody changed it.

Up to 10 in one chain, according to its own documentation; after that it gives up and treats the URL as an error. Browsers give up at roughly the same point. But the limit is not what matters: every hop is a round trip of latency paid by whoever visits the page, and Google passes less signal through each link.

Because each rule solves a different problem and nobody merged them: one goes from http to https, another adds the www, another adds the trailing slash. The fix is for every rule to point straight at the final URL in its canonical form, with https, www and slash, instead of at the next link. One redirect from any variant.

At least a year, which is what Google recommends for the signal to consolidate and for external links to stop pointing at the old URL. In practice, indefinitely: a rule costs nothing and removing it breaks the links left out there, in emails, documents and other sites that never get updated.

It is the way to say "this page was removed and is not coming back". Google de-indexes it sooner than with a 404, which it retries for months just in case. Use it for discontinued products, services that no longer exist and pages with no reasonable equivalent to redirect to. If they have one, redirect.

Less than years ago — Google eventually treats a long-lived 302 as permanent — but the process is slow and, meanwhile, the old URL remains the canonical one in its eyes. Across a whole migration that means weeks of rankings split between two URLs. If it is permanent, say so with a 301 and skip the wait.

From our server on Cloudflare, with an agent that identifies itself as EinscubeBot. It has to be that way: a browser cannot read the headers of another domain. We do not store the URLs you check; the result is cached for ten minutes so the same URL checked twice in a row is not downloaded twice.

Because the request leaves from a server of ours, and a server that agrees to fetch any address is a door into other people’s internal networks. Only public URLs on standard ports are checked. For a password-protected staging environment, use your browser’s developer tools: the network tab shows the same chain.