I welcome the appearance of data on in-browser PDF Viewers, resulting from #2212

Different browsers handle pdf in different ways, and the various PDF viewers have different capabilities, and this can have an impact on accessibility conformance.

AFAIK, only pdf.js used by Mozilla makes an effort to communicate PDF content to the accessibility tree (i.e. the subset of the DOM which is communicated to assistive tech). It makes a translation from PDF/UA tags to ARIA roles and attributes.

If people are deciding to use or not use PDF-in-browser on the basis of caniuse data, I believe they should be informed of different levels of accessibility support.