W3C home > Mailing lists > Public > public-tracking@w3.org > February 2012

Re: ACTION-110: Write proposal text for what it means to "not track" (ISSUE-119)

From: Roy T. Fielding <fielding@gbiv.com>
Date: Mon, 13 Feb 2012 15:59:33 -0800
Cc: Ninja Marnau <nmarnau@datenschutzzentrum.de>, "<public-tracking@w3.org> (public-tracking@w3.org)" <public-tracking@w3.org>
Message-Id: <C15702B4-AAD7-4328-8055-363B3FD18095@gbiv.com>
To: Nicholas Doty <npdoty@w3.org>
This is regarding ISSUE-5

On Feb 13, 2012, at 3:32 PM, Nicholas Doty wrote:
> On Feb 13, 2012, at 3:04 PM, Roy T. Fielding wrote:
>> On Feb 13, 2012, at 1:09 PM, Nicholas Doty wrote:
>> 
>>> Hi Roy,
>>> 
>>> On Feb 13, 2012, at 12:49 PM, Roy T. Fielding wrote:
>>>> Please be aware that this would require Apache httpd to respond
>>>> that it is always tracking, by default, regardless of how the
>>>> underlying services are implemented.  Likewise for Squid, TrafficServer,
>>>> haproxy, and all other HTTP servers that I am aware of.
>>>> 
>>>> If we can't find a definition that allows HTTP access logs and normal
>>>> retention for fraud control, then let's give up.  I will not implement
>>>> DNT if it can be used as a bypass for fraud and security controls.
>>> 
>>> As I believe Ninja noted, this is *not* intended as a set of requirements for compliance with a DNT header, just a meaningful and entirely optional description that a site can use if it absolutely isn't tracking.
>> 
>> I do not believe that is helpful.  It implies that anything in that
>> list is tracking, which is false, and it implies that any site doing
>> those things can't claim it is absolutely not tracking, which is not
>> a desirable result (it makes this standard useless).
> 
> I'm confused, why would that make the standard useless? I thought the group had largely agreed on compliance as a broad prohibition of tracking with an enumerated list of exceptions for business purposes where tracking is allowable.

No, the group has not even remotely agreed to any definition of tracking.
It wasn't even close -- five different ideas, each with distinct groups
interested in that idea.

> If a site doesn't need any of those exceptions and simply isn't retaining data about users, why would it be unhelpful for them to have a way to say so?
> 
> Maybe you're concerned about the terminology of "absolutely not tracking" as opposed to 'complying with the DNT preference'? Better terminology would be great. Personally, I just honestly can't see why it's harmful for sites to be able to say that they're going above and beyond a compliance standard.

It is harmful to place that text in the specification because there
are many other sites that are absolutely not tracking and those
sites will also be sending a response that says they are absolutely
not tracking.  Adding text in the spec that excludes those sites
is a contradiction in terms.

>>> If there is an alternate definition that could accommodate common httpd configurations and still communicate to the user that to a more complete level no tracking is occurring, it would be great to see that option.
>> 
>> Here is an alternative:
>> 
>> A party may claim that it is not tracking if
>> 
>> 1) the party does not retain data from requests in a form
>> that might identify a user except as necessary to fulfill that
>> user's intention (e.g., credit card billing data is necessary
>> if the user is making a purchase) or for the limited purposes
>> of access security, fraud prevention, or audit controls;
>> 
>> 2) when user-identifying data is retained for purposes other
>> than to fulfill the user's intention, the party maintains
>> strict confidentiality of that data and only retains
>> that data for a limited duration that is no longer than is
>> necessary to accomplish that purpose, thereafter destroying
>> or otherwise clearing the user-identifying data; and,
>> 
>> 3) the party does not combine or correlate collected
>> user-identifying data with any other data obtained from prior
>> requests, user-identifying profiles, or data obtained from
>> third parties unless specifically directed to do so by the user
>> (e.g., when a user initiates a login request) or for the limited
>> purposes of inspection for access security, fraud prevention,
>> or audit controls.
> 
> Is this alternative just a re-statement of one outcome of the compliance doc or do you think this is an optional level beyond compliance? (I believe we're aiming for the latter in ISSUE-119.) I personally would think "absolutely not tracking" wouldn't include retaining identifying data for business purposes outside of the user's intent for an indeterminate length of time.
> 
> Thanks,
> Nick
Received on Monday, 13 February 2012 23:59:57 UTC

This archive was generated by hypermail 2.3.1 : Friday, 21 June 2013 10:11:25 UTC