It isn't that hard and the cost is nothing
compared to the benefit. All things are
not strictly business decisions. Some are
matters of human values. If we believe XML
is not to be subject to such values, that it
is instead, strictly a tool of business, then
XML should and must always be subordinate to
SGML, a better design created by smarter
people for a customer that understood completely
the phrase "Good fences make good neighbors."
A choice is provided. Choose according to
your values. Do not let others choose for you
unless the options are of equal value and neither
has a significant discriminator.
As all too often, I am left looking at someone talking about the theoretical glories of the "fancy" version of something, and not able to disagree with the words actually said, but still thinking of my real experiences in the real world, that just don't match that theoretical glories being described.
In particular, I was "almost there" back in the 90s: I didn't work for Softquad, I was "pop" at a Mom and Pop CDROM operation. Real software was incredibly expensive. The SGML handbook cost a week's groceries. The language used at least appeared deliberately obfuscatory. In order to get any useful work done it was immediately necessary to override the default SGML declaration, which was clearly pointed at the s370 system of 10 years before. For all the beauty in the concept, there were also plenty of places where it sure looked to me like this bunch of document people hadn't bothered to learn lessons that comp sci language people had learned from COBOL and LISP. And it wasn't like there weren't books to teach these lessons.
Against that, the greedy self interest of the w3c members has led them to write specs that are comprehensible to the people who will implement them, make them readily available, make elucidating documents readily available, and to attempt to merge the accumulated lore of all the communities involved. To me these are all very important things, all of them missing from the ISO effort.
I simply don't find the choice as simple as you apparently do.