On Tue, 30 Mar 2004, Ernest Cline wrote:

> Explicitly using HYPHEN, NON-BREAKING HYPHEN, and MINUS
> handles the HYPHEN-MINUS ambiguity when you know how you
> want it the line to break around it.

In practical terms, they seriously limit the number of browsing situations
where the user sees the correct characters at all - or hears them. Not to
mention all kinds of other software than browsers. I recently learned that
Google translator cannot even handle a right single quote correctly,
and it is a much better supported character than e.g. the minus character.

In principle, on the other hand, using non-breaking variants of characters
and special characters for mere line breaking control looks primitive if
you compare it with the simple idea of markup. Unicode really tries too
much in this area. (It's comparable to language tag characters.)

> For most other isolated cases,
> ‌ and ‍ (ZWNJ and ZWJ) are sufficient.

Sufficient for creating great confusion, for sure.

I don't see why authors should use poorly supported tricky characters
instead of simple markup that has worked for years,

> You'll have
> to come up with a better example to convince me that <nobr> is
> needed for semantic reasons.

The practical reasons alone should be overwhelming. And in principle,
handling line breaking - especially to prevent line breaks inside a
string, as opposite to preventing it in a particular location between two
characters - belongs much better to markup level than to character level.

-- 
Jukka "Yucca" Korpela, http://www.cs.tut.fi/~jkorpela/