The most important of these non-paragraph blocks of text is verse. We'd *at least* like to see verse blocks to be unambiguously identifiable by machine processing, with some rule to tell authors "if you want to include some poetry, this is what you do...".) This will be useful for improving visual presentation options, will enhance accessibility (that is, the verse can be presented to the listener as verse and not as generic prose in text-to-speech engines) and may even aid more advanced searching mechanisms (which can be focused to search only in particular structures such as verse, for example.)
Most verse can be distinguished from prose by the preponderance of capitalization at the beginnings of lines, and the preponderance of punctuation at the ends of lines, which heuristics work until you meet e e cummings. But then, it has always struck me that the goal of zml is to meet 90% of your needs, 90% of the time. -- RS