Use language preferences in determining locales

[jessealama]

As discussed at the 402 monthly call on 2026-08-06, allow the set of initial locales in browser contexts to be determined by (1) the user agent as well as (2) language preferences (in other words, navigator.languages).

closes #1010

[eemeli]

This introduces a normative reference on the HTML spec, which should be noted in the relevant section.

[jessealama]

This introduces a normative reference on the HTML spec, which should be noted in the relevant section.

The Overview section where we're adding the new reference is non-normative, so I think it's OK to cite HTML here. We don't (currently) have something like Informative References, so the only place to add it would be the Normative References section, which I think wouldn't be right given the whole Overview section is to be understood as non-normative.

There's also one other reference to HTML in 402, in spec/locales-currencies-tz.html, about DefaultLocale. It's pretty much the same idea that the overview talks about:

The returned value is is a potential fingerprinting vector. In browser environments, it should match navigator.language to avoid providing any additional distinguishing information.

[eemeli]

The Overview section where we're adding the new reference is non-normative, so I think it's OK to cite HTML here.

Huh, didn't notice that. Given that the relevant paragraph already now includes this:

Furthermore, dynamic changes to these sets must not result in users becoming distinguishable from each other.

we maybe should restructure things so that this does become normative. While I am aware that we don't use RFC 2119 keywords, an expression like "must not" does sound awfully normative. And it should be.