Timing is everything in communications.
Spoken language has spaces between the words and yet they vary in length to signify or delimit clauses and other structures. A hyphen is almost like a half space, a comma is a double space, a semicolonial is ..., etc.
The problem is the weird little marks are restricted to integer multiples of time but the actually spacing is closer to a floating point variable. Maybe we need something more like music annotation to exactly specify word cadence and spacing. Otherwise every sentence sounds robotic.
I met someone involved in folk music who was trying to document Swedish folk songs. They were forced to use 1/128th notes to attempt to approximate the timing patterns in the songs.
How about we create a set of 4-bit vari-pause marks that can represent anything from 1/4 of space (one of Paul's nested short-hyphens) to a 4X space for a full stopper. Or should it be a log scale?