Thanks to Cyrille and Lars for their explanations.
I think I understand the desired solution, but I
wonder if (in the meantime) something useful might
not be achievable at the PS level (at least for
Latin 1 users). Specifically, rather than worrying
about the inability to query the PS interpreter
about stringwidths, how about forcing a string
width (e.g., via awidthshow) on the PS interpreter.
The result might be quite reasonable as long as the
initial width estimate isn't absurd.
fwiw,
Alan Isaac