fix grammar for "Vary" header field

[reschke]

it uses

"*" / #vary

but "vary" is a token, and token allows "*" anyway. Also, see #7.

[reschke]

In particular, "*" is a valid field name. We probably should reserve it in the registry.

[annevk]

(CORS has an issue like that as well, including for a * method; see the note at https://fetch.spec.whatwg.org/#http-new-header-syntax.)

[reschke]

Proposal:

vary-by = "*" / field-name
Vary = 1#vary-by

Then state that if "*" appears, no "real" field-name is allowed to appear as well (for backwards compat).

Do we need to discuss cases like

Vary: *, *

?

[royfielding]

I don't see why that is an improvement.

[mnot]

I thought we were going to reserve * and define its meaning in prose...

[mnot]

I.e., it should just be

Vary = 1#field-name

[reschke]

which implies registering "*" as reserved field name, right?

[mnot]

Yes. That's #274

[mnot]

Discussed; will create a PR to move ABNF and prose to list-based.

[reschke]

Do we need to list that as change from 7230?

[mnot]

Probably a good idea; will commit to master and link here.