I can only do math while on the clock and I’ve realized that also stands for complicated bug reports/feature requests/whatever this is exactly. Meaning I’m likely to miss something, but I’ll take a haphazard crack at it.
Anyway, my understanding of Markdown is that the content between square brackets is the alternative text and that captions aren’t explicitly universally supported. See Markdown Guide.
Lemmy-UI renders alt text from the content between square brackets, which is just perfect. Reddit, for what it’s worth, renders that - where it is supported - into alt text and a caption. That does lead to terrible alt text, where people (Reddit admins, specifically) only see the caption and forget where else that text is going. So yeah, having a separate caption sounds like a great idea.
That being said, you (or OOP?) may enjoy getting into the weeds of accessible name calculation. Will a screen reader read alt or title? Both? In which order? (VoiceOver on macOS read alt and then title. Who knew!?)
Now, I can’t imagine interface icons missing alt text will have anything at all to do with Markdown rendering. In fact, it’s possible they don’t need alt text at all, if the functionality is already accessible some other way, like through an adjacent text link. You can’t just leave out the alt attribute, you should hide the node, but I haven’t looked into kbin’s interface to find out if that’s the case. In fact, I haven’t looked into how kbin renders Markdown for images.
So… does any of that help? Does it help formulate specific questions, perhaps? Who should I address the bill to? Hehe.