[omega] Question about the paper published in EuroTeX 2005

[Chris Rowley]

I may have missed some of this `conversation' so I hope these ideas
are still pertinent.

1.  Some years ago, Yannis published a useful analysis of `ligatures'.

There he introduced (I hope I have the right names):

  mandatory ligatures

  aesthetic ligatures.

In terms of the current discussion I would say that the `mandatory'
ones are those that apply to `characters', whereas the `aesthetic'
ones depend on the font used and on the typesetting tradition; and so
these latter can be viewed as applying to glyphs or (much the same)
as being part of the glyph-choice procedure.

2.  It is probably better to continue to analyse glyph choice at the
level of words (or, rather, sub-words if ligatures should be
avoided at certain points (eg between parts of compound words).

Thus each `word' has many different possible `visualisations': some
of these variants depend on the fonts used, others on the
typographic tradition in use, etc.  Some of these should only be used
for `special cases' eg when a word is divided between two lines at a
particular point.

3.  Thus it is, for example, the job of a `paragraph formatting'
module to find enough feasible visualisations of the input text
(character string) of that paragraph and then to choose a `good
enough' visualisation (or maybe offer a choice to some other process,
in a more complex system).

This will require knowledge about feasible visualisations of
(sub)words (including punctuation) and their metric (and perhaps
other) properties and about the spaces between the words Plus a few
other layout specs).

3a. When using a font resource whose rendering engine must be accessed
via `sequences of Unicode slot numbers' there will be an extra step
needed in order to deliver the correct sequence to the font resource
(and to ensure that the right settings (for use of ligatures etc etc)
are used when the font resource interprets that sequence).  This seems
to me to be a peculiarity of this week's technology, so I am not sure
that this step should be part of a good general model of
characters/glyphs/@@@emes.

4.  I realise that this particular analysis, using the concept of `a
word', holds only for alphabetic scripts (and possibly some syllabic
ones) but I believe that the ideas can be extended to cover other types
of script.

5.  Some of this may be relevant to what the `objects' used by the
model (and implementation) are called.  Another thing that may be
relevant to choosing a name is that similar objects (ie with extensible
property lists) and structures (eg `network graphs') will be needed at
all levels: characters/glyphs, words, lines, paragraphs (in their many
forms), columns, table entries, tables, pages, spreads, ...

Note also that these do not form a nice tree-like hierarchy.

That's enough for now; maybe some more when I have read through the
interesting details of the many messages.


chris