Re: [http-state] Is this an omission in the parser rules of draft-ietf-httpstate-cookie-21?

[Adam Barth <[e-mail]>]

On Tue, Feb 15, 2011 at 12:29 PM, Roy T. Fielding <[e-mail]> wrote:
> On Feb 14, 2011, at 8:42 PM, Adam Barth wrote:
>> On Mon, Feb 14, 2011 at 7:23 PM, Roy T. Fielding <[e-mail]> wrote:
>>> Parsing for user agents is defined in section 5.  Servers have to
>>> parse cookies as well, and the grammar provided in section 4 is wrong
>>> for both generation and parsing.
>>
>> I agree that the grammar depicted in Section 4 is not appropriate for
>> parsing.  That grammar is not for parsing.  It's for generation.
>> Perhaps we should add instructions for how servers should parse the
>> Cookie header?  That's absent from the current document.
>
> We don't need more instructions.  We need a correct ABNF that tells
> us what to generate and what to parse.
>
>> As for generation, there's no exogenous notion of correctness.  We can
>> speak about usefulness, if you like, but correctness is endogenous to
>> this document.
>
> Then you should delete the introduction, since it says
>
>   This document specifies the syntax and semantics of these headers as
>   they are actually used on the Internet.  In particular, this document
>   does not create new syntax or semantics beyond those in use today.
>
>>> Why do you want to push forward a document with a known error in the
>>> grammar?  Just fix it.
>>
>> I disagree that it's an "error."
>
> Then please explain to Amazon why you want to break their site?
> Look at your browser's cookies for amazon.com and you will probably
> find cookies named session-token, at-main, and x-main that do not
> follow your grammar.  They are quoted strings and valid under all
> prior descriptions of the Cookie and Set-Cookie header fields.
>
> And while you are at it, maybe you'd like to explain to Google
> why you want to make Google Analytics cookies invalid.
>
> http://code.google.com/apis/analytics/docs/concepts/gaConceptsCookies.html
>
> The cookie named __utmz is not a quoted-string, not a token, and not base64.
>
>>> This is not a trivial detail.  As written, the specification says
>>> server developers SHOULD break sites like amazon.com.
>>>
>>>   cookie-value      = token / ""
>>
>> Can you explain how the developers of site B using a restricted
>> grammar for generation will break amazon.com?  That seems entirely
>> non-sensical.
>
> Site B is Amazon.com.  You are editing a proposed standard for the
> Internet and, last I checked, they use HTTP on the Internet.
>
>>> The correct grammar is in the original Netscape cookie spec: "This
>>> string is a sequence of characters excluding semi-colon, comma and
>>> white space."  Which easily translates (conservatively) into
>>>
>>>   cookie-value      = %x21-2B / %x2D-3A / %x3C-7E / %x80-FF
>>>
>>> and it does not conflict with section 5.
>>
>> I've updated the draft to recommend the grammar below for generating
>> cookie-values:
>>
>> cookie-value      = token / *base64-character
>> base64-character  = ALPHA / DIGIT / "+" / "/" / "="
>>
>> Do you have a particular use case in mind for generating Set-Cookie
>> headers outside this grammar?
>
> Yes.  Backwards compliance with existing content stored within
> several hundred million deployed browsers that won't expire
> until sometime in 2031 dictates that the grammar be
>
>  cookie-value      = %x21-2B / %x2D-3A / %x3C-7E / %x80-FF
>
> because servers have to parse Cookie in every form that they
> ever sent a Set-Cookie in the past.

You really think we should recommend that servers use invalid UTF-8
sequences as cookie-values?  That sounds like bad advice...

Adam


> Why are you arguing against this?  You are adamant about
> supporting the above syntax in browsers, for the same reasons,
> and there is no harm in supporting the complete Netscape syntax
> in the server grammar.
>
> ....Roy
>