Some journals will publish a separate file in XML, specially if they are open access. As for why PDFs, I think the answers are:<p>1. P (portable): you can access it offline, store it locally and annotate (something that researchers do extensively for key articles)
2. design: despite all possibilities with html, including html5, flex, interactive graphics and apps, pdf still allows you to have a much more granular control over design, and looks matter
3. tradition - old habits die hard<p>Specifically in relation to markdown, although some of its flavors are semantic, and that it also allows for absolutely awesome things such as R markdown (Rmd), tagging specific article sections is still complicated and cumbersome. Including div tags inside markdown has a number of issues. This is all to say that although there is a lot of people writing articles in Rmd, it is not an ideal solution.