Thank you for your answer and for thinking about it.
It would add flexibility and it would be almost complete for a lot of things.
Big last thing would be vertical merging of cells in a tab, and your app could be used
in scientific world.
Fletcher, if MMD is generating a figure element then the attributes specified in the markdown should be added to the figure element and not the img element within the figure.
It's not possible to properly style the image/figure if that is not done.
Nowadays, HTML5 document styling is done by assigning classes (via the "class" attribute) to HTML elements and using those classes to assign CSS styles from a style sheet. This keeps the HTML clean (no style blocks or verbose style attributes) and allows the powerful CSS selection capability to fully control document styling in a flexible and efficient way.
If I specify a "class" attribute in the markdown, and that class attribute is assigned to the img element rather than the containing figure element, then I have no way of using CSS to style some figures differently that others. And the img elements in every figure are constraint to however the figure element is styled globally.
If the "class" attribute gets assigned to the figure then I can use CSS to selectively style both the figure element and the contained img element, using different styles for different combinations of classes.
I think that one change would resolve most of the headaches around this issue and would help avoid embedding HTML in the markdown.