Re: ISSUE-45 ACTION-246: draft proposal regarding making a public compliance commitment

Trying to embed a choice of "compliance regime" in a header is a futile project.

(1) A server cannot know what legal regime it is operating under in an interaction with a user.  It can guess, but it may guess wrong.  It can try to choose, but that choice will not always be honored by authorities.  Formalizing the regime creates traps for the unwary; this could backfire badly on websites.
(2) Frequently, more than one regime will apply.  E.g., the same interaction may take place under California and federal law and NAI principles.
(3) What is required to "comply" with a given regime is itself a matter of tremendous uncertainty.  There will be skews between "USA" and what the FTC considers sufficient.
(4) Multiple implementations of the same "compliance regime" will differ materially in their interpretation of its terms.  Witness the extensive debates on this list.
(5) The complexity of standardizing and communicating the specifics of these different regimes increases the complexity of the standard by an order of magnitude. It is an open-ended commitment to the equivalent of a fresh Tracking Compliance and Scope document for each applicable compliance regime.  Perhaps this group will not have to do the work of drafting each version, but someone will.  And then there is the problem of managing the list of tokens: that implies a process for determining when a new one is warranted and adjudicating conflicts between multiple claimants.

Compliant participation in Do Not Track should involve a public commitment to a common baseline, defined by the Tracking Compliance and Scope standard.  The details beyond that will vary by jurisdiction, by industry group, by company, by site, and sometimes by user.  It is a hopeless effort to try to formalize these variations in advance.  Perhaps with time and experience some patterns worth exposing to users will become apparent and would justify an iteration to express some (but hardly all) of those patterns in an extended Tracking Preference Expression specification.  But these are not version 1.0 features, and compliance regimes are the wrong abstraction for expressing them.


On Sep 5, 2012, at 7:22 PM, David Wainberg <<>> wrote:

Hi All,

There are lots of assumptions and accusations flying around in this discussion which are not helpful.  Can we please back up for a second?  This proposal arose out of a working group call two weeks ago, during which I pointed out the potential legal landmine introduced by the notion of making the WKL in and of itself a public commitment of compliance with the W3C DNT standard. This led to a discussion of the diversity of actors that will be attempting to implement DNT, and the fact that there will necessarily be relevant distinctions in how they implement. I took an action item to propose an approach that better accommodates this diversity without creating a legal landmine. This is all this is -- an idea for dealing with the two issues that were identified on the call.  We are -- and continue to be - working on this in good faith and nothing has changed about anyone's commitment to anything.

Notwithstanding concerns about complexity and user confusion (which I'll address), it is my personal opinion, as I've already stated, that the current proposed language will inhibit wide adoption of the full spec. I think we're all agreed that the desirable outcome is to have something reasonable that will be widely used.

Regarding complexity and user confusion, I believe that users may not understand the W3C DNT spec any better than they understand privacy policies. Lawyers and engineers are going to have a difficult time determining exactly what the spec says can and cannot be done. In almost all cases, I believe that users' understanding will come from what's presented in the UA's UI and in the media.  It does concern me that we have almost zero control over either, and I expect that given where the spec is going, there will be a considerable delta between users' ideas about what DNT means and what DNT actually means. I fear this may be largely due to the fact that we're stuck with the term "track," which we are unable to define in a way that is consistent with the WG's policy aims and the common understanding of the meaning of the word.

Given that, the TPWG is making choices on users' behalf based on what the TPWG thinks should be reasonable, regardless of what common understanding is. So let's be clear that what's important is that in any case DNT means something in the ballpark of what a reasonable user would expect. This can be accomplished in many ways, and as long as it is reasonable, it's not going to make things substantially more confusing. The confusion will stem from the way the choice is represented to users.

So, this is the problem we have to solve. We need a standard that is reasonable for everyone, including the businesses that will be affected by it; that is relatively consistent with common understanding and expectation; that is not unnecessarily rigid and does not become calcified; and that will be widely adopted. My proposal was intended with those aims in mind.


On 9/5/12 12:06 PM, Shane Wiley wrote:

I believe this proposal and the strong support within IRC during the working group call would officially declare this as NOT a dead end.  It would be helpful to gauge the working group as I believe you’ll find considerable support for a compliance flag within the well-known location resource.

- Shane

From: Aleecia M. McDonald []
Sent: Wednesday, September 05, 2012 9:00 AM
To:<> (<>)
Subject: Re: ISSUE-45 ACTION-246: draft proposal regarding making a public compliance commitment

Of note: in Seattle, we discussed the possibility of having multiple codes to indicate different flavors of DNT. Specifically, I raised it as a suggestion. The WG members soundly rejected, in favor of coming to a common single understanding of DNT. We have already declared this a dead end.

One can imagine a world with, say, a DAA approach and a W3C approach, without needing a new flag sent with every response. Just pick different semantics. It will be very clear which is what, without the overhead. If that is the problem you are trying to solve, I think it is already solved without needing any work here.

If we take this just as being about different regions, I'm not sure what a USA or NLD designation entails. And I'm not sure how to convey that to users. I think I do not understand what you have in mind yet. I look forward to hearing more about how you think that could work.


On Sep 4, 2012, at 5:51 PM, David Wainberg <<>> wrote:

This fulfills ACTION-246 (, which relates to ISSUE-45 (

There are problems with the current proposed approach to issue 45. The current version does not accommodate implementation distinctions based on, for example, geography/jurisdiction, business model, or technology. It also creates unnecessary and counter-productive legal landmines that will spur companies to avoid implementing the full spec. We can provide for making legal commitments without this unwanted result.

I think the first point should be obvious. There will be a tremendous diversity of organizations, business models, and technologies to which DNT may be applied, either voluntarily or compulsorily, under a diversity of regulatory regimes. The spec needs to accommodate this diversity.

The more important point is that, if we make the mistake of tying the server response (the header or WKL) to a broad, legally-binding representation that goes well beyond the specific meanings of the responses, end-users will lose out because companies will avoid implementing the response mechanisms. The reality is that companies who may otherwise be eager to implement DNT will avoid making representations that could be construed in overly broad ways, that may be ambiguous, or that otherwise are potentially misaligned with what they do. Instead, companies will seek to make representations that unambiguously describe their practices. We should facilitate this, not make it difficult.

Note that I am definitely not saying that companies should be able to act contrary to what they represent in the response mechanism(s). That, however, is not a problem we need to solve. Companies will be held to account for any such misrepresentations anyway, regardless of what the spec says. And if the available responses are sufficiently precise and adequately defined, I think companies will implement them.

This proposal solves both problems. It will provide for the enforceable statement that the working group is aiming for, but it will also allow needed flexibility for servers operating under various regulatory regimes, and would do so especially for servers operating under multiple regulatory regimes. And, most important, it would create a mechanism whereby companies can clearly and accurately say what they do and then do what they say.

The proposal is the following:

  *   The compliance spec remains silent on the matter
  *   Add a required "compliance" field to the tracking status resource in the TPE, where the value indicates the compliance regime under which the server is honoring the DNT signal.
  *   The value of the compliance field is a 3-5 letter token indicating the applicable regulatory regime. Allowed tokens could include 3-letter country codes, e.g. USA, GBR, NLD, or designations for voluntary regimes, e.g. W3C, DAA, NAI, IABEU. My understanding is that an organization like IANA can manage a list of tokens in order to prevent collisions.

Received on Thursday, 6 September 2012 01:31:09 UTC