- From: Jonathan Mayer <jmayer@stanford.edu>
- Date: Tue, 2 Oct 2012 13:09:13 -0700
- To: "Roy T. Fielding" <fielding@gbiv.com>
- Cc: "public-tracking@w3.org (public-tracking@w3.org)" <public-tracking@w3.org>
- Message-ID: <1070B1E849C042EA8D8F87D2196F96F2@gmail.com>
Roy, What's the "rabbit hole" this text spares us from? In my reading it roughly tracks where the Compliance document currently stands, save considering service providers to be part of the first party. Jonathan On Tuesday, October 2, 2012 at 8:46 AM, Roy T. Fielding wrote: > The Compliance spec is overly focused on details that aren't relevant > to the user's expectations regarding tracking. I'd like to fix that, > but I haven't been able to piece together enough focused writing time > to deal with Compliance given my other obligations on TPE. So, this > is a brief sketch on a way out of this particular rabbit hole. > > = = = = = = = = = = = = = = = = = = = = > > Each user action on the Web, such as selecting a link or > submitting a data form, can result in numerous protocol requests > (interactions) by the user agent to satisfy that action. These requests > might be sent to a single HTTP server, to multiple servers owned by > the same entity, or to a variety of servers owned by multiple > entities. > > This specification distinguishes between "first party" sites > (origin servers owned or controlled by the entity that the user > expects to interact with as a result of their intended action) and > "third party" sites (all other origin servers). Unfortunately, > there is no programmatic way to distinguish the two, since the > user's expectations are often determined by the context in which > an action is presented. > > For example, a user might select a button containing the logo of > a familiar company, but which actually links to an advertising referral > counter, which redirects to an auditing service, which also > redirects to a page at the familiar company's site, which in turn > instructs the user agent to retrieve a number of page elements, > which might include elements from the original referral site, > the advertiser, the auditor, the familiar company, and any number > of domains operated on behalf of that company. A user agent > will not know which of those sites are considered the first party, > since it cannot know why the user chose to select the logo. > > Similarly, origin servers cannot control all of the ways in > which their resources might be linked from other sites. A server > might receive requests for page elements that were never expected > to be used by other sites, and thus receive data about user activity > that it had no intent to process. > > However, resources on the Web are usually designed for specific > purposes and linked to within specific contexts. In general, it > is possible for origin servers to design a resource for use in a > first party, third party, or dual-use context. In particular, > resources that are intended for the purpose of tracking user > activity across multiple sites are almost always sensitive to the > context in which that activity occurs. > > In other words, this specification does not rely on a strict > technical definition to distinguish first party resources > from third party resources. Instead, it relies on implementations > of resources that are intended or predominantly used in a third > party context to comply with the requirements on a third party, > and for implementers to ensure that tracking resources designed > only for first party use are either discouraged from being used by > third parties or constructed such that tracking is disabled when > used in a third party context. > > The scope of a "first party" is determined by user expectations > and control over the data collected, not limited to a specific > site, domain, or legal entity. The first party includes the legal > entity that owns and controls the site intended by the user's action > and any employees, officers, affiliates, and contractors that operate > on behalf of that entity and that are bound under contract to keep > data collected on behalf of that entity confidential within the > scope of the first party, separated from data collected on behalf > of other parties, and to have no independent rights to use, share, > or retain the collected data except as directed by the first party. > A site that is operated on behalf of multiple legal entities is > considered to have a joint first party if a user's expectation > would be that each of those entities have control over the data > collected. > > = = = = = = = = = = = = = = = = = = = = = = = = = = = = > > Cheers, > > ....Roy
Received on Tuesday, 2 October 2012 20:09:40 UTC