W3C home > Mailing lists > Public > public-p3p-spec@w3.org > July 2005

Re: Trouble with data schema

From: Lorrie Cranor <lorrie+@cs.cmu.edu>
Date: Thu, 7 Jul 2005 07:41:16 -0400
Message-Id: <286ec6e4aa78cf0a3f182982494fac24@cs.cmu.edu>
Cc: 'Rigo Wenning' <rigo@w3.org>, 'public-p3p-spec' <public-p3p-spec@w3.org>
To: Giles Hogben <giles.hogben@jrc.it>

Ah, ok, I understand now. I think saying we cannot create new 
categories is a reasonable limitation for P3P 1.1. Nobody has ever done 
this or expressed interest in doing this, although I understand 
theoretically why someone might want to do it in the future. But I 
think reving to a new version of the data schema in the case that there 
was consensus for a new category makes sense.

As for the question of whether category names are elements or not.... 
is there a downside to making category names an element other than 
having to change the transforms?

As far as impact on APPEL, as long as the old format is included in the 
policies, APPEL can continue to use that if need be. At some point 
APPEL needs overhaul, and then it would make sense to switch to 
evaluating the new format, so then I think it could be made to work 
with either elements or not, so I don't think that's a worry.

Lorrie


On Jul 7, 2005, at 3:57 AM, Giles Hogben wrote:

>
> The change I'm suggesting is that we say new schemas cannot create new
> categories, they can only refer to categories from the base data 
> schema.
> That is they can say that their elements are in categories but they 
> can't
> make new categories and put the elements into them.
>
> It would be even better, given time, to convert the BDS categories 
> (which
> are currently expressed as string values of CATEGORY tags, for xsd 
> reasons)
> to data types like all the rest, but standing at the top of the 
> hierarchy
> (as they are semantically speaking). That means that elements using
> categories would then look like
> <demographic><dynamic><cookie/></dynamic></demographic> instead of
> <dynamic><cookie><CATEGORY>demographic</CATEGORY></cookie></dynamic> - 
> but
> this would change the transforms and as you say open up a can of worms.
>
>
> -----Original Message-----
> From: public-p3p-spec-request@w3.org 
> [mailto:public-p3p-spec-request@w3.org]
> On Behalf Of Rigo Wenning
> Sent: Wednesday, July 06, 2005 18:41
> To: 'public-p3p-spec'
> Subject: Re: Trouble with data schema
>
>
> again, a list problem...
>
> ----------  [fwd]  ----------
>
> Subject: [Moderator Action] Re: Trouble with data schema
> Date: Wednesday 06 July 2005 16:48
> From: Lorrie Cranor <lorrie@cs.cmu.edu>
> To: Giles Hogben <giles.hogben@jrc.it>
> Cc: 'Rigo Wenning' <rigo@w3.org>, public-p3p-spec@w3.org
>
> I still am not getting what you mean by disallowing categories in new 
> data
> schemas. As rigo said, the whole point of categories was so that user 
> agents
> would have some clue about new data schemas. I'm not really sure what 
> change
> to the current working draft you are proposing.
>
> Lorrie
>
> On Jul 6, 2005, at 10:38 AM, Giles Hogben wrote:
>> I think disallowing categories in new data schemas does not open up a
>> can of worms. It probably closes up a few.
>> Changing the XSD we already have does... (and would be quite a lot of
>> work).
>>
>>
>> I don't fully follow what you are proposing... but I am strongly
>> opposed to any changes that open up nasty new cans of worms at this
>> point in time.
>>
>> Lorrie
>>
>> On Jul 6, 2005, at 6:03 AM, Giles Hogben wrote:
>>> Hmmmm - I didn't mean that you can't keep using the old categories. I
>>> just mean to simplify the syntax, let's forbid them in future. The
>>> point is that
>>> in any new schema, broad categories could simply be expressed as
>>> e.g.
>>>
>>>
>>> <navigation><cookie/></navigation>
>>>
>>> OR
>>>
>>> <demographic><vehicle><color/></vehicle></demographic>
>>>
>>> One possibility would be to offer them as elements in the new XSD BDS
>>> so that the weird syntax string based ones disappear completely in
>>> new schemas
>>> and you just use the element to hook new data elements onto.
>>>
>>> We could even turn the XSD schema up on its head and get rid of the
>>> string based categories so that the XSD defines the categories as
>>> elements with the
>>> category would then be at the top level in the data element and
>>> details
>>> would be hooked on below.
>>>
>>> E.g.
>>>
>>> <navigation><cookie/></navigation>
>>>
>>> You could also write them without categories unless they were dynamic
>>>
>>> <user><online><uri/></online></user></navigation>
>>>
>>> Agents would match elements like <navigation/> to all the allowed
>>> subelements of <navigation>
>>>
>>> This would mean quite a lot of rewriting the transforms etc... I
>>> estimate 2 weeks extra work. It may also have nasty implications for
>>> APPEL. Any thoughts.
>>>
>>>
>>> -----Original Message-----
>>> From: public-p3p-spec-request@w3.org
>>> [mailto:public-p3p-spec-request@w3.org]
>>> On Behalf Of Rigo Wenning
>>> Sent: Wednesday, July 06, 2005 11:28
>>> To: Giles Hogben
>>> Cc: 'Lorrie Cranor'; public-p3p-spec@w3.org
>>> Subject: Re: Trouble with data schema
>>>
>>> Am Wednesday 06 July 2005 09:27 verlautbarte Giles Hogben :
>>>> I would like to propose a simplification which there is still time
>>>> to put in the spec. We include categories in the Base Data Schema
>>>> for backward compatibility - but we disallow them in custom schemas
>>>> from now on.
>>>
>>> I don't get you here. One of the features of the category system was,
>>> that new custom data elements could be attached to the existing broad
>>> categories thus giving the custom elements some meaning. This meaning
>>> would then be easier to understand for user agents.
>>>
>>> ***There is nothing to stop you doing that.
>>>
>>>> If you want to put in broad categories, there's nothing stopping you
>>>> - you just have to make them into data elements which subsume the
>>>> narrower categories. There's no need for a completely different and
>>>> confusing syntax.
>>>
>>> Can you give an example how this would look like?
>>>
>>>> Yes it's because of XSD. Basically because each
>>>> data element takes a subset of categories from a global set, this is
>>>> only possible with a custom data type (or at least that's the only
>>>> way we could find to do it and we consulted some XML lists)...
>>>
>>> Can we say that we don't allow for NEW categories? Because in your
>>> example in the Spec, you say:
>>>
>>> <element minoccurs="0" maxoccurs="1" name="musical-preference">
>>>   <element minoccurs="0" maxoccurs="1"
>>> ref="classicalmusic-preference"/>
>>>     <annotation>
>>>       <documentation>
>>>         Musical Preferences
>>>       </documentation>
>>>     </annotation>
>>>     <element ref="CATEGORY" minoccurs="0" maxoccurs="*"
>>>              type="allCategories"/>
>>>   </element>
>>>
>>> So if you wanted new <category> - Elements, you would have to make
>>> them available in a custom XML Schema? How would that fit in our
>>> framework?
>>>
>>> Best,
>>>
>>> Rigo
>
> -------------------------------------------------------
>
>
>
>
Received on Thursday, 7 July 2005 11:42:19 GMT

This archive was generated by hypermail 2.2.0 + w3c-0.30 : Thursday, 7 July 2005 11:42:20 GMT