structure:true without aria attributes

[pkra]

It would be very helpful not to have it simultaneously to allow the exploration of other patterns. It also avoids having to clean up the attributes to work around SRE idiosyncrasies.

[zorkow]

For clarification: You want the skeleton attribute only, without any other aria annotations (like owns)?

[pkra]

For clarification: You want the skeleton attribute only, without any other aria annotations (like owns)?

Sorry, I used the wrong terminology -- I'll update the title.

No, I don't want a skeleton attribute.

I'm looking for whatever the structure: true, settgin was producing before the recent addition of aria-* and role attributes -- IIRC those were just data-semantic-* attributes but maybe I'm missing something.

From a quick look through recent commits, I'm guessing I'm looking to disableaddAria from fadf19d6bd55ed3a1f1e8b7b53df7078c8d97713. (The generated aria stuff is also problematic in itself but that's a separate issue).

[zorkow]

That's what I meant. And I probably knew that this issue was coming when I opted against YAO... But I can hide it behind YAO! No problem, but I am running out of letters. Any suggestions?

[pkra]

Any suggestions?

I'd say stop using ARIA attributes.

First, I think ARIA markup belongs on the mathjax end of things - they determine user facing behavior which is not what SRE usually provides.

Second, they are weirdly implemented; you seem to have followed my advice half-way with problematic results (both here and in the client-side walker).

Third, developers (like myself) who want to use alternative patterns now have to remove these first which is cumbersome and risks dropping actually useful information.

[klobetime]

ARIA does provide user-facing behavior, but arguably so does SRE -- the entire point of generating speech from math is to present it to the user in some fashion. I agree that ARIA can be problematic, but removing it entirely seems overkill. At the very least, aria-label is extremely useful in correcting issues that can't/shouldn't be corrected in other ways, such as in #632.

Not sure that @pkra was actually advocating removing everything ARIA related rather than hiding behind a flag, but wanted to make this point just in case. :)

[pkra]

Not sure that @pkra was actually advocating removing everything ARIA related rather than hiding behind a flag, but wanted to make this point just in case. :)

I see how this might have been confusing. To hopefully clarify: I requested that the option structure:true goes back to its old behavior - not adding any ARIA attributes. (SREs enrichment output is then still present in the output using data-semantic-* attributes.)

To explain again: the ARIA markup currently produced with this option is invalid and forces authors into a very specific ARIA pattern. Neither seems like a good idea.

(I realize now that #632 appears to be about input, not output so I may have misunderstood your point entirely.)

[zorkow]

PR #706 adds a new flag aria which is set to false by default. So you should no longer see the aria attributes.