>> The only real solution is to completely redesign the layout engine,
>> switching to a TeX-like paragraph handling algorithm.
>
> I admit to not knowing the internals of the formatting algorithm,
> but I don't see why solving this problem requires the algorithm to
> change. To solve it, internally groff could just do what it
> currently advises the user to do: pretend the line is longer by the
> amount of padding being added to each glyph. (groff already knows
> this amount; to the user, it's not obvious, as the amount is
> calculated based on parameters the user supplies.)
For backwards compatibility you no longer can change the behaviour of
`.tkf', so this would be a *new* command (or maybe `.tfk' with a fifth
parameter).
> Granted, the issue is somewhat academic, as revamping the formatting
> algorithm is already considered a higher priority than tweaking
> .tkf, and there's no one available to do either one.
This is true, unfortunately. However, given that groff is written in
fairly clean and conservative C++ with quite a good object class
hierarchy, implementing a new formatting algorithm shouldn't be too
difficult for people who like the object-oriented approach.
Werner
