This is a great illustration of the concept: https://web.archive.org/web/20060613193727/http://diveintoma...
If true believers can make this kind of mistake, how often will it happen to people who are just trying to get work done? People will have bugs in their code sometimes, but what should the failure mode be?
So again, I understand that it's nearly impossible to make the transition now, but that doesn't mean it wouldn't be preferable. And there's no reason the transition couldn't still be made, but much more slowly, and perhaps without ever making the big browsers reject by default.
I don't know if there are other disadvantages to XHTML, but if the strictness issue is the only one, then it seems like there would still be value in slowly transitioning over.
The analogy is a classic example of the "Everything must act like a compiler" fallacy.
There are many situations in which compilation is hopelessly inappropriate as a user model.
Not only is the web one of them, but the hypothetical semantic web is also one of them.
You can't force semantics into compilable tokens. The suggestion that you can - and should - is nonsensical.