I hear you on needing piles of CLI packages. Regarding the web browser, the pandoc HTML seems pretty basic, so even the lightweight GUI browsers will probably handle it correctly. Also, if you include the correct bits in each image link in the markdown, I believe the resulting HTML5 will render with your included image descriptions in a text-only browser.
Here[1] is the HTML 5 standard. It includes a lot of things that a word file would never need. You're correct about not needing a lot of the things that say a browser like chrome has. Can we still not achieve the same thing with something simpler?
