Cookies and Accept-* interactions

[thpts]

^{(@mnot requested a new issue during Thursday's httpbis meeting at IETF 103)}

#549 suggests the idea of an algorithm to to key the cache on a cookie by name.

One consideration in this specification should be the use case where some value in a cookie can override server Accept behaviours. As a known use case, websites that offer multiple languages may allow a user to set a cookie which determines which language. When that value is not present in a cookie, the server falls back to using Accept-Language or some other alternate.

I think this specification should perhaps make some clarity over Cookies doing this kind of override, even if it may fit into the Vary section (as servers implementing this probably are using Vary: Cookie)

[mnot]

Thanks!

[mnot]

I think this would require some sort of "all" value (*?) in Variant-Key; if we had that, it should be fairly straightforward, I think.