thin_space causes collateral damage in eqn

[Douglas McIlroy]

man eqn says
         set thin_space n
causes n/100 em space to be "automatically inserted after punctuation
characters", which are vaguely defined as "characters such as ',' "

I used thin_space 0 to take away automatic space after . in formulas.
To my surprise it also took away intentional spaces represented by ^.

The loss of ^ threw away a lot of carefully crafted spaces, and even
threw away spaces in some sporadic cases where I wanted to override
thin_space 0.

Another parameter, thick_space, has the same effect on ~.

Apparently thin_space serves double duty--to control automatic spacing
and to set the width of ^.

This coupling of function strikes me as a bug. One thinks of ^ and ~
as groff keywords, not characters. And even if they are characters,
they represent white space placed deliberately, not "automatically
inserted".

Doug