[texhax] Awful looking output

Uwe Lück uwe.lueck at web.de
Wed Oct 14 13:20:24 CEST 2009


At 06:30 13.10.09, Oleg Katsitadze wrote:
>On Mon, Oct 12, 2009 at 01:13:32PM -0400, Michael Barr wrote:
> > Try this:
> > $HW=WH$
> >
> > Don't TeX's kerning tables usually do a better job than this.
>
>Well, that's the kerning for the math italic font.

At 02:44 14.10.09, Reinhard Kotucha wrote:
>On 13 October 2009 Michael Barr wrote:
>  > All this is true, but the point was that you should not put math formulas
>  > in \mathit (which is for italic text) or, at any rate, you should not 
> have
>  > to do so.  It is arguable that it was the WH that was wrong since there
>  > was not enough space and it looked like text, not math mode.  Whatever,
>  > TeX is supposed to work out the spacing automatically and doesn't.
>
>No.  Kerning tables have nothing to do with TeX at all.  They are font
>specific.  TeX uses whatever is in the .tfm files, regardless where
>they come from.

Are all of you sure that TeX uses any kerning table in $HW=WH$. Please look 
at Appendix G of the TeXbook, "Generating Boxex from Formulas".

My impression is that TeX does /not/ do any kerning here -- because it 
would be a bad idea. Kerning is good within words and only there. Kerning 
between two letters that are to represent a product or a composition would 
be confusing, making the misleading impression the two letters formed an 
/atomic/ name with an own meaning, similar to, e.g., "mod" or (in category 
theory) "ker".

On the other hand, I think there /is/ something awful here: the way TeX 
deals with slanted symbols and fonts. Same problem with $\overline{H}$ and 
that you sometimes have to deal with italic corrections (setting them with 
plainTeX or avoiding them with LaTeX). TeX treats a slanted symbol as a 
/box/ (|box|), a /rectangle/ (|rectangle|). This is why $WH$ is wider than 
$HW$ (slanted "W" has its leftmost "dot" at its top, "H" at its bottom). 
Wouldn't it be nicer if it treated it as a /parallelogram/? 
(/parallelogram/ indeed!) Trying to set them as close as possible?

Cheers,

   Uwe.



More information about the texhax mailing list