However, I changed the format when I saw Oleg Kiselyov's SXML work, to make my Web-scraping and other tools able to use his XML tools. I later made a few other tools that used SXML, such as a simple HTML-writing template embedded in Racket that does some of the checking and work at compilation time.[1]
At a library level, SXML's arbitrary nested lists make some things computationally harder to do (e.g., find all the attributes, depending on the "normal form" of SXML), but some other things easier to do (e.g., some kinds of functional editing, due to arbitrary nested lists). Aside from those considerations, SXML is the closest to a de facto standard for XML and HTML tools in Scheme.
If I ever happen to have funding to do so, I'd like to revisit the exact representations, to try come up with an end-all-be-all for all purposes, and redesign/reimplement all the tools from scratch. Until then, there's SXML.
[0] some people can't bear lisp uniformity for instance.
I bring this up so often that I have a go-to blog post for it: http://chriswarbo.net/blog/2017-08-29-s_expressions.html