Feedback from Freedom Scientific: single vs. multi-attribute approach

[ggordon-vispero]

We favor the single-attribute JSON model. If a multi-attribute approach is ultimately chosen, we're going to end up packaging up all of them into a single structure anyway in order to pass them to our speech rendering subsystem. So being able to retrieve them as a single string feels more direct. It also does a better job of isolating these attributes into their own little walled garden separate from the other attributes of an element.

Chrome and Firefox use the IAccessible2 interface by default which has a way of providing a screen reader with all exposed attributes as a single string, so retrieval will be similar in both the single and multi-attribute cases. Of course they'll need to allow these attributes through before screen readers can easily test, since only select attributes are made available via IA2.

UI Automation on Windows does not currently have such a mechanism. Each attribute currently needs to be exposed as a specific UIA property. In this case, the single attribute approach would mean that only one new property would need to be exposed. UI Automation is always cross process which means that all other things being equal, the fewer function calls to retrieve things the better.

[alia11y]

Thank you for your feedback, Glen. We will wait to hear the feedback from other AT vendors.