JAWS 17 should not make any changes regarding the buttons-- it’s supported aria-label since at least 2010/JAWS12.
A note about keyboard shortcuts: be aware most screen readers form a barrier between a user’s keystrokes and the browser, so that “k” can take me to the next link or “h” can bring me to the next heading. A screen reader will hijack these keys and for good reason. It’s why keyboard shortcuts built in Javascript (and using different keys for each and every site since we don’t have any unified standards, except that they seem to be vaguely vim-like at times) are hotly debated still in standards circles and not really considered accessibility enhancements, but power user enhancements. Just because your Javascript is listening for a keystroke does not mean a screen reader has even bothered to pass it on to the browser or your Javascript in the first place.
In any case, this may be more an example of someone like that person we all know who has a lot of trouble using something like Gmail. Gmail is “technically accesisble” but it’s built for nerds and power users. I have trouble with it (guess I’m not nerd enough), because it takes the basics everyone wants (an email client should let me read mail, write mail, delete mail, forward mail, and add attachments) and overcomplicates it by adding in a buttload of feature-creeped options on top that at least I am always accidentally triggering some way or other and it just fights me all the way. However, technically, it does not discriminate against users because they have a disability.
Discourse is the same thing. It’s a glorified forum software that loads so much more fluff and stuff on top that it’s restricted to certain kinds of users. It does not progressively enhance, it assumes all its users are nerds with the latest and greatest of hardware, software, and network. It assumes their icons always load and that all people magically know what they mean (ug). This is a deliberate decision on Discourse’ part, so things like usability and accessibility considerations must be taken in that light. Also keep in mind that both Marco and Nolan are nerds too. They are developers who also happen to use screen readers, plus Marco specifically works in the accessibility space. Neither are representative of Joe User who also happens to use a screen reader. They’ll catch actual bugs, and can offer some judgement calls, but that doesn’t mean if all those issues were fixed that anyone using JAWS will effortlessly be able to figure out how to use Discourse software.
Yeah, I personally am of the “Tim Berners-Lee definition of accessible” which means progressive enhancement, working reasonably regardless of hardware/software/someone’s income level/education/disability/whatever. But the official definition is something like “does not unduly hinder or lock out users simply because they have a physical disability.” That is the definition Discourse wants to use, meaning many people looking for forum software should look elsewhere.
Lynx can work fine with any basic forum that uses forms with POST to add content. Any such forum can load Javascript that, when it runs, HIJAXes those forms into quicker AJAX requests and whatnots. Discourse does not discriminate against “visually impaired” users by not being a text-based-but-still-mostly-text forum. It discriminates against “text browser” users (among others). Even the WebAIM surveys (which yeah are self-selected and therefore probably over-represent webby nerd people who use computers regularly in all ways) show most users use the same graphical browsers non-SR users use (with less Chrome use of course).
My unsolicited advice to the Discourse devs is to continue to have users of AT (and keyboard, and touch, and speech, and and and…) continue to test Discourse and give feedback to devs, and not close issues like TechnoBear’s without fixing them, since anything that is continuously developed can continuously add new bugs you didn’t have back when, for example, Marco did some testing. It’s part of your regular QA, right? And developers are testing for accessibility as they build any new thing, for example by using an automated tester like Tenon.io right?
This will help Discourse improve and maintain its technical accessibility. And obviously any random AT user can find a bug and should report it, that’s good. But I don’t see Discourse working for stuff like Lynx any time soon, and I can’t say that counts here as accessibility issues. If there are JAWS bugs, then yes those are bugs. The email thing sounds promising, though.