I'm not sure this is right- what motivated this change?

I ran into it while testing the dialog with NVDA where it doesn't read out the content at all. I based it on the example from MDN.

In MDN's example, the element they're pointing to is literally a description of the dialog, though. In reality, that element will contain all sorts of arbitrary content, so it's not really accurate to say it's the describedby for the dialog element. We should support setting describedby on the dialog, but not automatically grab the entire content for it.

AFAIK the content of the dialog would be available just like any other content on the page when navigating with a screenreader. How were you moving through the dialog after opening it?

I tabbed my way to one of the buttons and opened the dialog. The dialog header got read out correctly, as well as the button, but none of the content. I also tried tabbing and clicking around, but I couldn't get it to read out anything. FWIW, VoiceOver works as it is in master.

Typically you would use NVDA's navigation to go through text content (vs. jumping between form fields with tab).

http://webaim.org/resources/shortcuts/nvda

You're right, I retried it and it got read out this time. I'll re-purpose the PR to add a config option for aria-describedby.

Done.

This issue has been automatically locked due to inactivity.
Please file a new issue if you are encountering a similar or related problem.

Read more about our automatic conversation locking policy.

_{This action has been performed automatically by a bot.}