[twobyfour]

That seems to me like a rather anglocentric view. How are people who speak Arabic, Hindi, or Greek supposed to use the internet? And believe it or not, not everyone's keyboard allows for easy input of ASCII characters.

Besides, it's too late. There are already non-Latin TLDs out there - in part to support languages with non-Latin alphabets.

[Meai]

I'm the first guy to defend the need for multiple languages but how then make it safer? Maybe highlight which characters are unicode in the url string...

[twobyfour]

That's a good point. There's probably a way to highlight unexpected characters. I wouldn't use "Unicode" as a criterion (that's an encoding, not a character type that excludes ASCII). Perhaps the browser could define a set of "expected" characters based on the user's locale and highlight anything outside that set.