@MostlyBlindGamer @stefan The distinction between ALT and descriptions is a relic of HTML 4.01 and does NOT apply to social media.

HTML 4.01 contains both a terse descriptor—the ALT attribute—to give a verbal thumbnail of the image, & a verbose descriptor—the LONGDESC element—to provide a rich, formatted descriptions which can contain MathML, SVG, & other markup necessary to actually provide an equivalent experience for the blind/VI & those for whom supplemental speech enables them to make sense of a complex image, equation or table.

Mastodon only provides ALT, so it must serve as a hybrid descriptor. A mere caption is NOT sufficient.

You may wonder why I cited HTML 4.01 & not HTML5. Over my formal objection, as well as the formal objection of many others, LONGDESC was deprecated in HTML5.

Of course, deprecating an element does not mean that that element is no longer necessary. Both a terse & a verbose descriptor MUST be part of any ML's native semantics in order for the dyad to function correctly and provide us with the information we need.

ALT in HTML 4.01/XHTML 1.0 was envisioned as a thumbnail, to provide the user with an informed opinion about whether it was worth it or not to follow a link, or to request more detailed information about the image. On Mastodon ALT is the only means to provide an image description that is programmatically bound to the image it describes.