Re: [Devel] Kerning problem with capitals

[Werner LEMBERG]

> Freetype FAQ states that Freetype doesn't do kerning/positioning,
> period.

Correct.  It only gives access to the simplest kerning format, as a
convenience function.

> Phylosophically, I believe that Freetype is the place to deal with
> generic word rendering.

I disagree, and I've explained my reasons already.  Am I alone with
this opinion?

> This is because freetype deals with fonts, which may or may not
> contain kerning info.  Freetype can also use autokerner
> (rsb_delta+lsb_delta) in conjunction with autohinter.

But rsb_delta + lsb_delta isn't an autokerner at all!  It just
corrects a possible glyph displacement which happens due to rounding
advance widths within the autohinter at small ppem values.  Use
ftstring and display `LT' in Vera.ttf (9pt, 96dpi), and cycle through
the kerning modes with forced autohinting: You can see that there is
no `rsb_delta + lsb_delta' correction, but still a kerning value is
applied.

> I have to repaet, that because of autohinting, there must be auto
> adjustment in spacing.  Freetype simply knows what happens to glyphs
> and even can evaluate hinting quality.  It's way too much
> postprocessing to be passed on to higher level applications for a
> simple word rendering.

To summarize: Two corrections must be applied for small ppem values if
the autohinter is used: `rsb_delta + lsb_delta' and kerning.  The
former could indeed be handled by FreeType, the latter can't in
general.  As mentioned in a previous mail, I don't object if the
tutorial contains such a function for pedagogical purposes (probably
based on the code from ftstring.c).


    Werner