Issues that this engine needs to handle:

 - Picking the right character based on language/character
 - Composition of combining diacriticals - should use precomposed
   diacriticals when possible
 - Finding word/line/breaks

For each sequence of character + combining marks

 - Find combined representations from most combined to least combined.
 - If there are particular characters for this combination, use
   them, otherwise 

==========

Ligature table per font.
Pick fonts in order of preference according to LANG

Characterstics of western/basic:

 Consistent LTR
 No reordering
 Many characters => one glyph
 Character/Word/Line breaking dependent only on character properties

Issues
 
 Accent placement for Greek, Vietnamese
 How do we handle a case where lang == en_US, we have a glyph
   combination with a combined form in iso-8859-2 but not
   in iso-8859-1?