URL inspect

The parse itself is the browser’s URL interface, because that is the WHATWG algorithm and therefore the definition of what a browser will do with the string. Writing a second parser would only create disagreements with reality. Everything else here exists because the URL interface cannot answer it.

The query is re-parsed here rather than handed to URLSearchParams, which treats ?a and ?a= as the same thing — one has no equals sign, the other an empty value, and some back ends distinguish them — and which cannot tell you a parameter appeared twice. ?id=1&id=2 means different things to different frameworks (first wins, last wins, becomes an array), and collapsing it hides the reason a request behaves oddly. Every pair is kept, in order, duplicates included.

The host appears in two forms, because an internationalised domain has two: the browser connects to the Punycode xn-- form while the address bar shows Unicode. RFC 3492 is implemented here in both directions, and while it is at it the tool flags any label mixing two of the three look-alike alphabets — Latin, Cyrillic, Greek. That is the standard homograph trick: replace the a in apple with a Cyrillic а and no reader can tell.

The tracking-parameter list comes from each ad platform’s own documentation and Firefox’s query-stripping list. Any such list expires: networks add parameters faster than lists are updated, and any site is free to make one of them load-bearing, which is why it is editable on the page. The path, by contrast, is never touched — not re-cased, not reordered, not normalised. /A and /a are different resources, and a cleaner that tidies them breaks links.