Better document various TrueType hinting modes.

Text based on private communication with Greg Hitchcock and added to the
repository with his permission.
This commit is contained in:
Werner Lemberg 2015-09-10 07:44:48 +02:00
parent e5a27a7596
commit b6eef5ae7b
1 changed files with 77 additions and 0 deletions

View File

@ -52,6 +52,83 @@ FT_BEGIN_HEADER
*
* The TrueType driver's module name is `truetype'.
*
* We start with a list of definitions, kindly provided by Greg
* Hitchcock.
*
* _Bi-Level_ _Rendering_
*
* Monochromatic rendering, exclusively used in the early days of
* TrueType by both Apple and Microsoft. Microsoft's GDI interface
* supported hinting of the right-side bearing point, such that the
* advance width could be non-linear. Most often this was done to
* achieve some level of glyph symmetry. To enable reasonable
* performance (e.g., not having to run hinting on all glyphs just to
* get the widths) there was a bit in the head table indicating if the
* side bearing was hinted, and additional tables, `hdmx' and `LTSH', to
* cache hinting widths across multiple sizes and device aspect ratios.
*
* _Font_ _Smoothing_
*
* Microsoft's GDI implementation of anti-aliasing. Not traditional
* anti-aliasing as the outlines were hinted before the sampling. The
* widths matched the bi-level rendering.
*
* _ClearType_ _Rendering_
*
* Technique that uses physical subpixels to improve rendering on LCD
* (and other) displays. Because of the higher resolution, many methods
* of improving symmetry in glyphs through hinting the right-side
* bearing were no longer necessary. This lead to what GDI calls
* `natural widths' ClearType, see
* http://www.beatstamm.com/typography/RTRCh4.htm#Sec21. Since hinting
* has extra resolution, most non-linearity went away, but it is still
* possible for hints to change the advance widths in this mode.
*
* _ClearType_ _Compatible_ _Widths_
*
* One of the earliest challenges with ClearType was allowing the
* implementation in GDI to be selected without requiring all UI and
* documents to reflow. To address this, a compatible method of
* rendering ClearType was added where the font hints are executed once
* to determine the width in bi-level rendering, and then re-run in
* ClearType, with the difference in widths being absorbed in the font
* hints for ClearType (mostly in the white space of hints); see
* http://www.beatstamm.com/typography/RTRCh4.htm#Sec20. Somewhat by
* definition, compatible width ClearType allows for non-linear widths,
* but only when the bi-level version has non-linear widths.
*
* _ClearType_ _Subpixel_ _Positioning_
*
* One of the nice benefits of ClearType is the ability to more crisply
* display fractional widths; unfortunately, the GDI model of integer
* bitmaps did not support this. However, the WPF and Direct Write
* frameworks do support fractional widths. DWrite calls this `natural
* mode', not to be confused with GDI's `natural widths'. Subpixel
* positioning, in the current implementation of Direct Write,
* unfortunately does not support hinted advance widths, see
* http://www.beatstamm.com/typography/RTRCh4.htm#Sec22. Note that the
* TrueType interpreter fully allows the advance width to be adjusted in
* this mode, just the DWrite client will ignore those changes.
*
* _ClearType_ _Backwards_ _Compatibility_
*
* This is a set of exceptions made in the TrueType interpreter to
* minimize hinting techniques that were problematic with the extra
* resolution of ClearType; see
* http://www.beatstamm.com/typography/RTRCh4.htm#Sec1 and
* http://www.microsoft.com/typography/cleartype/truetypecleartype.aspx.
* This technique is not to be confused with ClearType compatible
* widths. ClearType backwards compatibility has no direct impact on
* changing advance widths, but there might be an indirect impact on
* disabling some deltas. This could be worked around in backwards
* compatibility mode.
*
* _Native_ _ClearType_ _Mode_
*
* (Not to be confused with `natural widths'.) This mode removes all
* the exceptions in the TrueType interpreter when running with
* ClearType. Any issues on widths would still apply, though.
*
*/