This page summarizes potential accessibility-related use cases of the Reporting API identified by the Research Questions Task Force.
The Reporting API operates directly in the user agent context to provide performance-related data to Web applications, which are delivered on a "best effort" basis. The specification establishes a framework for this reporting; specific report types defining the meaning and format of the data transferred may be specified elsewhere.
The Research Questions Task Force is considering what information it may be appropriate to include in one or more potential report types that could be created to convey data relevant to issues of accessibility to the operators of Web sites and Web applications.
Whether a person has a disability or uses assistive technology is a private matter that the individual may not wish to disclose. Therefore, any reporting which only operates in the presence of an assistive technology or which is only active for users with disabilities needs to be subject to strict privacy controls, such as voluntary, informed, prior consent.
List of Possible Use Cases
Note: these use cases have not been prioritized. This is an idea collection.
- Reporting of reliably identifiable accessibility-related errors on Web pages that can readily be discovered by the user agent. Examples include t he absence of captions in multimedia, and the absence of alternative text for images. These reports would provide evidence of issues associated with pages and media resources that users are actually visiting, in contrast with a sample of pages on a site that may be taken for Web accessibility evaluation purposes, or a report on the entire site generated by an automated evaluation tool.
- Reports of accessibility-related issues identified by the user agent that occur in third-party content embedded in Web pages that may be subject to frequent change, may be dependent on the user's geographical location, and may thus be inconvenient to audit reliably with Web accessibility evaluation tools.
- Error reports from machine learning models run in the user agent context. For example, image recognition or speech recognition confidence estimates could be reported. How useful would this be in practice?
- Reports on environmental conditions (e.g., ambient light and sound) likely to affect the accuracy of image and speech recognition systems, or which may affect the quality of audio and video in real-time communication applications. Again, how useful would this be in improving application performance over time?
- Reports of accessibility issues flagged by the user while interacting with Web pages/applications. In the simplest implementation, the user could focus on an element or select part of a page and then activate a control to identify it as having an accessibility-related issue. In a more complex implementation, the user could write and submit a comment on the accessibility issue. (This would require support from the application, from the user agent, or from an extension, presumably.) This use case assumes that interactive reporting requiring user input is permissible.