If I request an URL like https://www.w3.org/ns/rdfa in the browser, it automatically loads the HTML version of that resource since the browser also sends Accept: text/html. If I request a different media type, such as via curl -H "Accept: text/turtle" https://www.w3.org/ns/rdfa, I get a different representation of the resource.
This shows that a single HTTP-based resource may in fact refer to/consist of multiple sub-resources. Preferably, these should all be "views" of the same resource, but they may not be ‒ the content that is loaded could differ by the specified media type, language, the current time, or any other header. While you can store the whole request in a format like WARC that replicates everything about it, you don't usually need to know every header sent to the server; only those that produce a different response.
Is there a URI scheme, notation, or another public or proposed standard that makes it possible to refer to these narrower "resources"? It could allow something like this:
https://www.w3.org(Accept:text/html)/ns/rdfa
https://www.w3.org(Accept:application/rdf+xml)/ns/rdfa
https://www.w3.org(Accept:text/turtle)/ns/rdfa
to identify the three versions of the page with a header that differentiates them, both for the purpose of expressing such a distinction, and for automatically setting the header when such a URL is navigated to. I chose to state the header before the URL path, because the request only contains everything from the first / after the hostname (up to #), and it makes sense to store it in a place that already affects the headers (hostname sets Host, credentials set Authorization).
Something that can express this specific distinction are web links, for example the <link> tag in HTML:
<link href="https://www.w3.org/ns/rdfa" type="text/html">
<link href="https://www.w3.org/ns/rdfa" type="application/rdf+xml">
<link href="https://www.w3.org/ns/rdfa" type="text/turtle">
But that only encodes the media type, and, of course, is not a single URI.