W3C has advanced the longdesc attribute to a Recommendation, overruling objections from browser makers.

Not a single browser vendor supported advancing this specification to recommendation.

Apple formally objected when it was a Candidate Recommendation and provided lengthy research and documentation (better than anyone has before or since) on why longdesc is bad technology (in practice has not and does not solve the problems it claims to).

Mozilla formally objected when it was a Proposed Recommendation, agreeing with Apple’s research and reasoning.

Both formal objections were overruled.

For all the detailed reasons noted in Apple’s formal objection, I also recommend avoid using longdesc, and instead:

• Always provide good alt (text alternative) attributes for images, that read well inline if and when the image does not load. Or if there’s no semantic loss without the image, use an empty alt="".
• For particularly rich or complex images, either provide longer descriptions of images in normal visible markup, or linked from a image caption or other visible affordance. See accessibility expert James Craig’s excellent Longdesc alternatives in HTML5 resource for even more and better techniques.

Perhaps the real tragedy is that many years have been wasted on a broken technology that could have been spent on actually improving accessibility of open web technologies. Not to mention the harrassment that’s occurred in the name of longdesc.

Sometimes web standards go wrong. This is one of those times.