Lossless --embed-resources (Respect html5 download attr)

[avidseeker]

example.html (converted from markdown using --embed-resources)

<img
	 src="data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAMIAAADIBAMAAABIVCNkAAAAD1BMVEX//wD///8AAAD/AADOzgCL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">

Running

$ pandoc -f html --extract-media img example.html

results in img/[SHA1].png which is expected.

However, html5 offers download attribute which offers a way to mitigate against filename loss when embedding resources. E.g: <img download="smile.png" src="..."> should be converted to ![](img/smile.png) instead of ![](img/[SHA1].png).

[silby]

I can't find any indication that the download attribute is standardized for use with <img>, only <a> (and <area>). If Pandoc grows the ability to round-trip the filenames of images through embedding and extracting, it probably would use a custom attribute (e.g. data-original-filename in the HTML output) rather than download.