WebAIM - Web Accessibility In Mind

E-mail List Archives

Re: Z-Order and Tag Order Need to Match?

for

From: Duff Johnson
Date: May 28, 2020 12:27PM

> My comments are prefaced by acknowledging that the majority of document authors still don't know how to or won't create well designed and structured digital content.

Acknowledged! :-)

> However, this week I received my second PDF with the PDF/UA Identifier on it that is not accessible. Headings were not tagged and paragraphs were broken into two paragraph tags whether they were in the middle of the page or at the bottom of one page and continued on the top of another. My previous tagged PDF with the PDF/UA Identifier on it had the entire first page in a list tag with the entire contents of the first page in a list item tag...Headings, paragraphs, lists, images...everything. Additionally this second document with the PDF/UA Identifier on it had empty nested paragraph tags three deep. I didn't think this was "allowed" in PDF or PDF/UA.

Of course it's not allowed. The document clearly fails, and is non-conforming. A good start would be to hold whoever applied the PDF/UA flag accountable for that decision.

…and that's really the purpose of the PDF/UA flag. It cannot itself assure compliance - anyone can add a mere flag. The flag represents an affirmative claim of conformance; it's an accountability device, not an accessibility one.

> NONE of the automated tools caught this including PAC 3. I thought at first that it may be because of the PDF/UA identifier that none of the automated accessibility checkers were finding these problems with the tags. I removed the PDF/UA identifier and ran the automated accessibility checkers again, including PAC 3. All told me that I had an accessible document when I could see that there were some structural/semantic violation in the Tags Tree. So, not the fault of an inexperienced remediator or someone who just runs the automated tools...they were told that the semantics were conforming.

These tools generally advise that the validation of the semantics (correct application of tags to content) is a so-called "human check" - up to the user of the validation software to evaluate. The job of a validation tool is to help the user perform this check… not to make it on their behalf.

> While it is true that document authors play an important role, maybe the most important as they create the digital content, and there are several tools that create tagged PDF, it is becoming more difficult to get clean Tags in PDF documents even if you start with an accessible source document. We are inundated with what I call garbage tags that simply bloat the Tags Tree with endless nested Tags that shouldn't be nested. As with the two previously mentioned documents, they weren't complex, even if I had used the auto tag feature in any of the conversion tools, I should not have the results I had (with the exception of the headings not being tagged, which is why you go down the Tags Tree or use the Order Panel to see what got tagged and what didn't).

This is very valuable input for PDF creation software developers.

> Since it is clear that we have adaptive technology that uses either the Tags Tree or the Reading Order, we need to have communication between those two panels to make sure that the adaptive technologies render the same reading experience no matter which panel they are using...not sure how we do that when we can't get a clean Tags Tree.

That "communication" is… Tagged PDF. As I've said earlier in this thread, the Reading Order cannot support semantics, and so its output cannot ever conform to WCAG 2.0 Level A!

Instead, developers need to simply support Tagged PDF.

> If we clean up the Tags Tree, would it then clean up the Order Panel information/architectural view of the document?

It's a fine idea for a feature-request. There would be some familiar complexities, e.g. from objects that mask each other, but the software could also ask for the user's help in resolving such issues, so I think it's a very fair thing to ask for.

> How do we influence the specs to get both working together so we can fix things in one panel and have them accurately reflected in the other??

The ISO committee looked at requiring alignment of the tags and content order years ago and concluded it wasn't possible in a wide variety of very common use cases (e.g. paragraphs that span pages) and therefore not viable as a PDF/UA requirement.

Duff.



>