Re: [Devel] Kerning problem with capitals

[Robert Clark]

On Tue, 2004-12-07 at 00:05 +0100, Werner LEMBERG wrote:

> I fear that the `bad' way is the correct one.  Here the details for
> the word `Red' (you can this easily verify with fontforge) if handled
> without anti-aliasing:
> 
>   glyph       left bearing   glyph width   right bearing
>   ------------------------------------------------------
> 
>     R              1              7              0
>     e              1              6              1
>     d              1              6              1
> 
> The advance width is 8 pixels for all three glyphs.
> 
> Kerning values:
> 
>   `R' + `e': -92 font units
>     -> Vera.ttf has 2048 font units per em
>     -> this is -0.539 pixel at 9pt at 96dpi (12ppem) -> -1 pixel
> 
> Horizontal coverage before applying kern values:
> 
>   012345678901234567890123
> 
>    +++++++ ++++++  ++++++
> 
> After applying kern values:
> 
>   012345678901234567890123
> 
>    +++++++++++++  ++++++
> 
> As can be easily seen, we have one pixel right bearing for the `e' and
> one pixel left bearing for the `d', which makes a two-pixel gap.  The
> one-pixel gap between `R' and `e' disappears after applying the
> kerning, and this makes the two-pixel gap stick out.

  So it's not just me that thinks this looks odd?

  I don't know much about fonts, so forgive me if this is a stupid
question: How does hinting (bci hinting in this case) interact with
kerning? I'm assuming that, since it changes the shapes of glyphs there
needs to be some adjustment to take account of this.

  The reason I ask is that the spacing looks much more natural with
hinting disabled (h key) in ftstring.

> It might be interesting to compare this with Windows, but I doubt that
> we get different results.  

  I did have a shot at this, but it looked so bad I gave up pretty
quickly. I probably don't know enough about configuring fonts in
Windows.

        Thanks,

                Robert