> that hired more people analysing these M$ formats than to develop their own product
Yes, yes. Run after moving target. There qwwas an article, describing in details why Microsoft every 2 years dropped its own standards and proclaimed next Great Silver Bullet :-)
Don't remember the link and anyway offtopic here.

1st is not trial and error. It is between "iterative approach" of maths and "worse is better" of programming :-)
You just settle the interface, and allow hoster plugins to implement it if they choose to.
Let them jump the wagon at their own pace.

2nd approach, "the Right Thing", sound good if you can "hire more programmers, than..."
Obviously this is not the case. That would never end. New hosters would appear more often, and old hosters would change mangling more often, than that proposed standard would ever be finished. Not saying about "implemented"

And if that sounds disgusting, then consider 1st approach as a mere field experiment, as a way to collect data for later systematic analysis :-)