Re: [websec] draft-ietf-websec-strict-transport-sec issue: "directive, name" and "directive value"

[Julian Reschke <[e-mail]>]

On 2012-07-10 00:42, =JeffH wrote:
>  > The following came up in my AD review of
>  > draft-ietf-websec-strict-transport-sec, and Jeff suggested that I
>  > needed to take it to the list.  So here it is.
>  >
>  > The ABNF in Section 6.1 has this:
>  >
>  >    directive = token [ "=" ( token | quoted-string ) ]
>  >
>  > Below that, bullet 3 says this:
>  >
>  >    3.  Directive names are case-insensitive.
>  >
>  > And in Section 6.1.1:
>  >
>  >    The syntax of the max-age directive's value (after quoted-string
>  >    unescaping, if necessary) is defined as:
>  >
>  > Nothing defines what a directive name or a directive's value is.  You
>  > and I know they're what's on the left side of the equals sign and the
>  > right side, respectively.  We can't assume, though, that people will
>  > figure out that the ABNF definition above turns into "name=value", and
>  > will thus know what those terms mean, completely unambiguously, for
>  > essentially all readers.
>
> fyi/fwiw, the manner in which the ABNF is crafted was finalized in the
> thread, with Julian Reschke, rooted here..
>
> Re: [websec] STS ABNF, was: new rev:
> draft-ietf-websec-strict-transport-sec-04
> https://www.ietf.org/mail-archive/web/websec/current/msg01114.html
>
>
>  > Nothing defines what a directive name or a directive's value is.  You
>  > and I know they're what's on the left side of the equals sign and the
>  > right side, respectively.  We can't assume, though, that people will
>  > figure out that the ABNF definition above turns into "name=value", and
>  > will thus know what those terms mean, completely unambiguously, for
>  > essentially all readers.
>  >
>  > Making the grammar like this will fix it:
>  >
>  >    directive = directive-name [ "=" directive-value ]
>  >    directive-name = token
>  >    directive-value = token | quoted-string
>  >
>  > If there's a good reason not to make the ABNF change above, I'm happy
>  > to accept some other way of defining the terms, but I think they must
>  > be defined.  I think doing it with the ABNF is the easiest and
>  > smoothest way.
>
> I can see doing it as above, or even as a comment..
>
>      directive = token [ "=" ( token | quoted-string ) ]
>                ; directive-name = directive-value
>
> Julian apparently has some reasoning for trying not to put everything
> into the ABNF (see the thread pointed to above).  So I think it'd good
> if he weighed in on this.
>
> I do note that the ABNF in draft-ietf-httpbis-p2-semantics-19 for the
> "expect" header field which Julian points at does explicitly define ABNF
> for expect-name and expect-value, similarly to Barry's suggestion above.

Adding ABNF productions for clarity seems fine to me.

Best regards, Julian