More thoughts below...
> Is there a reason that a document using DPUB roles can not use standard ARIA roles when they apply?
My understand is that the DPUB roles are complementary, so it’s not a choice of one or the other.
> In the case of "landmarks" role, the description perfectly matches the standard "navigation" role. So, why not just use role="navigation"?
I think that might be more a failing of the definition. In epub, the landmarks navigation is special in that it is required to be flat and every link must carry a semantic identifying the structure it points to. It’s sort of like declarative markup for special behaviours.
I’d be more inclined to find a way to make the list redundant, like merging with toc, than use a generic role like navigation. If we go that route, it seems problematic to figure out which navigation element to use.
But I’m relatively new to this work, so I’ll defer to Markus and Tzviya on whether this landmarks nav is expected to have the same function, whether we need more detail to make that happen, or whether there’s an alternative approach for the web general.
> It also seems that role="heading" would perfectly map to the proposed role="title".
Title allow the author to identify the component of the heading that represents the title, it’s not necessarily the heading itself (even if it sometimes does get applied to ranked heading tags). EPUB also has ordinal and label, for example, so that a heading can be broken out into constituent parts (e.g., for content remixing):
<h1><span epub:type=”label”>Chapter</span> <span epub:type=”ordinal”>1</span> <span epub:type=”title”>Into the Abysss</span></h1>
Title also morphed to identify the title for components that wouldn’t normally carry ranked headings (lists and the like). Without support for the outlining algorithm, there’s also a desire to identify titles in figure and blockquotes without using ranked headings and impeding on navigation by heading.
Another definition that probably needs some tweaking.
Matt