I think you are talking about different approaches here. Joost seems to propose the native path in analogy to the current implementation of Json.Decode, while Nick is in favor of having an Elm-only approach.
From what I read an Elm-only package would be a big thing to start. I see the following issues here:
(1) XML is tricky to implement correctly and I am not an expert here.
(2) There is no well tested Elm parser/lexer to build on as far as I know.
(3) From my experience, parsing XML is a task where performance issues arise quickly. So this should be considered from the start. To give an example, I intend to parse GPX files that may have a size of several MB with thousands of XML nodes.
In contrast, using native browser support gives you a tested XML parser with reasonable performance.
Correct me if I'm wrong here, but this is how I see it.