Re: [http-state] Minor things concerning draft-ietf-httpstate-cookie-01

[Adam Barth <[e-mail]>]

On Tue, Jan 12, 2010 at 3:44 AM, Ian Hickson <[e-mail]> wrote:
> On Tue, 12 Jan 2010, Adam Barth wrote:
>> On Tue, Jan 12, 2010 at 2:54 AM, Georg Koppen <[e-mail]> wrote:
>> >>> 5) Section 5 says that there are some algorithms specified for some
>> >>> tasks concerning the cookie protocol but the user agents do not have to
>> >>> follow these ("user agents are free to implement the cookie
>> >>> protocol..."). But, for instance in Section 5.1.1., there it says "The
>> >>> user agent MUST use the following algorithm..." Thus, am I free to
>> >>> implement it differently or not? It's a bit confusing...
>> >>
>> >> The operative requirement is:
>> >>
>> >>       <t>Although some parts of the cookie protocol are specified
>> >>       algorithmically, user agents are free to implement the cookie protocol
>> >>       in any manner as long as their resultant behavior is "black-box"
>> >>       indistinguishable from a user agent that implements the protocol as
>> >>       described.</t>
>> >>
>> >> Is there some way we can be clearer that user agents are free to
>> >> implement the protocol however they like?
>> >
>> > Why MUST the user use these algorithms? Wouldn't it be enough if the
>> > user MAY use them? I mean, they are free to implement the protocol
>> > however they want to do it, provided they come to the same results.
>>
>> The goal is to require that they come up with the same results.  I
>> guess we could say something more elaborate each time.  It just seems
>> a bit wordy.  Any suggestions for text to use?
>
> HTML5 uses:
>
> # Conformance requirements phrased as algorithms or specific steps may be
> # implemented in any manner, so long as the end result is equivalent. (In
> # particular, the algorithms defined in this specification are intended to
> # be easy to follow, and not intended to be performant.)

Thanks.  I've replaced the existing paragraph with this text.

Adam