Re: [http-state] Ticket 11: Character encoding for non-ASCII cookies values

[Adam Barth <[e-mail]>]

On Thu, Mar 4, 2010 at 12:23 AM, Daniel Stenberg <[e-mail]> wrote:
> On Wed, 3 Mar 2010, Adam Barth wrote:
>>> I read the section 4.1.1 and it says that servers "SHOULD NOT send
>>> Set-Cookie headers that fail to conform to the following". Where is the text
>>> saying that clients MUST accept all those other characters?
>>
>> It's a consequence of the parsing algorithm in
>> <http://tools.ietf.org/html/draft-ietf-httpstate-cookie-04#section-5.2>. In
>> particular, how the name-value-pair and the cookie-value are extracted from
>> the set-cookie-string.
>
> Yeah, that is in my eyes strikingly hard to comprehend and find info in when
> used as a reference. The step-by-step parsing algorithm is to me a very
> strange way to document a protocol syntax. Not the least because there's
> already ABNF syntax for 4.1.1 and even in 5.1.1.
>
> I'm not sure how I'd like to see this fixed but I would prefer an ABNF
> approach with added appropriate prose rather than the long free text
> algorithm in 5.2. We're documenting a protocol, and if that protocol can be
> implemented using other code flows then I don't think it is our spec's job
> to limit that as long as the protocol is followed.
>
> I'll try to write up a suggestion for this soonish.

Thanks.  I too would prefer a declarative presentation, but we need to
be careful not to sacrifice accuracy.

Adam