SoftwareHTTPweb historyprivacy
HTTP kept the misspelled Referer header, then spelled Referrer-Policy correctly
The historical quirk becomes a practical debugging trap when request fields, privacy settings and JavaScript sit beside each other.

The web can deliver an entire application to your pocket and still expects you to reproduce a spelling mistake to say where a request came from. Referer has one fewer r than the English word. Spellcheck would fix it. Your server may then have a completely different field on its hands, courtesy of the same helpful spellcheck that keeps trying to improve everybody’s surname.
RFC 1945, section 10.13, records the name in HTTP/1.0. The field lets a client identify the resource from which it obtained the requested address. That gives servers information useful for logging and link maintenance, with the rather substantial catch that the referring address can itself contain private information.
The modern HTTP semantics specification preserves the name and explicitly admits “the field name is misspelled.” There’s your typo, with standards-track paperwork attached to the name implementations already recognize and would have to change if somebody decided to improve its spelling.
Header names are case-insensitive in HTTP, so capitalization and spelling are separate issues. referer and Referer identify the same field. Adding an r produces Referrer, which is a different sequence of letters. A case-insensitive comparison still needs the same letters to compare.
The privacy control makes the trap easier to fall into. The W3C Referrer Policy specification uses Referrer-Policy for the response header that sets policy, and referrer for the HTML meta element’s name. Those spellings are deliberate parts of their respective interfaces too. Correcting everything with one search-and-replace would manufacture a bug while tidying up the code.
The HTML Standard’s Document interface adds the JavaScript spelling. document.referrer reports the document’s referrer as a string. A developer moving between a browser console and a request log therefore crosses the spelling boundary without leaving the same debugging session.
This is also where the headline-sized joke needs a limit. Using the wrong name does not inevitably crash a browser or produce a spectacular error. It can simply leave the application looking at a field that has none of the information it expected. Quiet failure is plenty annoying on its own.
That makes two checks useful together: whether the code uses the right interface name, and whether the browser should send the information at all. Treating every absent value as a typo would miss the privacy behavior. Treating a misspelled lookup as privacy protection would let a bug hide behind a feature.
Old interface names tend to survive because other software already depends on them. A new setting can get a clean name, while the old wire format keeps its historical wart. In this corner of the web, proofreading requires knowing which mistake to keep.
One r has grandfather rights.