It's like the author forgot about XML and the failures of its draconian error handling and decided to do it all over again.
It ignores the reality of implementations, and of commercial software development in general. Implementations aren't written in an ivory tower with an endowment for long-term maintenance, they're written in ten days against hand-wavy business requirements at companies that no longer exist in their original form. Users will go elsewhere, to software which uses formats and protocols which don't have draconian error handling.
Wishing it were another way won't make it so, it'll just give you XML all over again.
Some discussions from the time, none of whose arguments this RFC addresses:
https://web.archive.org/web/20081210125015/http://diveintoma...
http://www.aaronsw.com/weblog/001025
https://web.archive.org/web/20090416022813/http://diveintoma...
https://web.archive.org/web/20090309211058/http://diveintoma...