boundary char and implicit kerns

[Hilmar Schlegel]

Han The Thanh wrote:
> 
> Dear all,
> 
> I sent the following question to tex-implementors list, but didn't get any
> answer yet. In case anyone know the answer, I would be much appreciated.
> 
> ----- Forwarded message from thanh -----
> 
> Dear TeX implementors,
> 
> does anyone know why implicit kern is not treated as discardable? Under
> which circumstances an implicit kern can end up that the beginning or the
> end of a line after breaking a paragraph into lines?
> 
> It seems to me that it can happen for kerns with the boundary char. Is this
> the only case when it may happen?
> 
> Thanks in advance for the answers,
> Thanh
> 
> ----- End of forwarded message from thanh -----
> 
> In particular, I found quite unfortunate that when using a font
> containings kerns with the boundary char, character protruding (ie marginal
> kerning) stops to work for certain characters. These characters are
> preceded/followed by an implicit kern with the boundary char, which is not
> discardable. For example, the period or comma char are not protruded
> if there one has in an afm something like:
> 
> KPX period space 118
> KPX comma space 118
> 
> and space is set as the right boundary character.

Perhaps I misunderstand something here: isn't it exactly the other way
round that hanging characters (mid-line and/or begin/end-line) work in
Tex only with the above described configuration?
I.e. industry-standard fonts have kerns with respect to space (if at
all) to allow for optical kerning with resepct to word-boundaries. Using
space as a placeholder and translating into kerns with the Tex
boundarychar is an obviuos way to make use of this concept in Tex
despite it doesn't have a space character.

This works fine as far as I could see. Possibly you should consider how
you achieve this effect in another way. In any case are the kerns with
respect to the boundary character the first/last things which Tex sees
at both ends of a line. It is essential for hanging characters that
these kerns are not discarded.

Perhaps it is necessary to take those corrections in the AFM into
account when defining hanging in an alternative approach to see the
correct spacing at the end?

best regards,
Hilmar Schlegel

> 
> So now the question is: should implicit kerns be treated as
> non-discardable?
> 
> Regards,
> Thanh