[omega] Question about the paper published in EuroTeX 2005

[Chris Rowley]

Gábor

> Sorry to interfere with your dialogue

Fine ... We only do it in public so that people can interrupt!

> Chris:
> are you implying that a modern font technology today does not
> necessarily need to be based on a Unicode-to-glyph mapping?

Definitely.

> That "glyph
> access via Unicode" is just a consequence of OpenType's inability to
> access glyphs by their "names"?

A good question: I would answer that `I hope it does not have to be'.

It is not clear to me that the only purpose of a font is to `visualise
Unicode strings'; maybe this is all that is needed for typical
text-based applications right now; but there may be many ways (some
not yet thought of) other than `strings of Unicode slots' for
representing text in computers.  Also, many subtle aspects of
typesetting arise because it is doing far more than that; certainly
far more than when an operating system has to display a `text file' on
a screen.  the interesting question is whether all of this extra stuff
can be reduced to the same `Unicode slots'  that work well for the
basic screen display.

> I believe the OpenType people's choice to use Unicode-based glyph
> lookups was well considered. Correct me if I'm wrong but to me the term
> "glyph name", in the Adobe sense, suggests that there is a finite set of
> glyphs (typographic signs) that may be designed.

That was not the sense I was thinking of.  I was not thinking of
named-glyphs necessarily existing independent of a font but that the
designer of a font (or rather the packager of a font resource) should
give names to each glyph in that resource and advertise this name (it
need not be a human readable name but it probably will be as glyphs
are rather human things) so that typesetting software can, if
necessary, directly ask for a single glyph in that font `by name',
rather than having to know what string of Unicode characters will
force that font resource to typeset that single glyph.

Of course, now I have found a computer-readable naming scheme: just
name everything by the shortest sequence of Unicode slots that will
produce it:-).

Actually, that does not work because not every glyph can even be
accessed in this way.

Anyway, glyph names could be from some generally agreed list but need
not be.  If a typesetter wants to use Poetica then it would need to know
something about some naming scheme for Poetica in order to use it wisely.
It would be easier for a programmer if that scheme used human names
but numbers could be used.  Unicode slots (or strings of them) cannot be sensibly 
used by either humans or software when the glyph set is so far from
the character set.

> If a font allows glyph
> lookup only by names,

There is no requirement for this to be the only way to access glyphs.
Font resources where (mostly) there is only one visualisation
available for any character can use Unicode slots as names for glyphs.

> the typesetting engine will need to have a "glyph
> name list"

In my model the typesetter needs to be able to ask a font resource if it has a
glyph with certain properties, one of which could be a name from some
external list.

> characters. But what if a font designer would like to include a
> completely new glyph, let's say, a ligature, that has no official name
> yet? How could such a glyph be used by the typesetting engine?
> 
> The OpenType model (as well as AAT etc.) widens the possibilities of the
> font designer by giving control over how pure Unicode character streams
> become glyph streams.

That is good, indeed very good, for `non-typeset' material.

> The Unicode character set is finite by definition
> (at least at any given point of time; let's hope they won't go much
> beyond 20 bits in the near future...) whereas there should be no
> limitation of the glyphs that may exist.

That is one reason why they need names, not numbers.  But remember
that in my model a glyph does not exist outside a font resource.

> Establishing glyph names
> independently of Unicode slot numbers seems redundant to me at the best,
> and a severe limitation of the font designer's possibilities at the
> worst.
>

As I said, all that is happening currently is that the only names
available are `sequences of Unicode slot numbers'...and even those
have human names so those names (or sequences of names) would be
better than just the numbers.

Perhaps the ability to advertise what glyphs are available is more
important than them having human readable names, but I cannot see how
to advertise this without some kind of names.

> Or maybe I didn't get your point? What kind of glyph naming would you
> suggest that could avoid these problems and provide something more than
> Unicode-based access?

Any advertised means of accessing the individual glyphs in a font resource gives
more than Unicode-based access.  Having just Unicode-based access
means that the typesetter may never know what glyphs are available
unless it happens at random to send the font resource the correct
Unicode string.

Note that I am not saying that just having names is sufficient: the
typesetter may need to know, for example, what languages a particular
ligature was intended for, by the designer.  Maybe the font resource
provides different ligatures (of the same basic glyphs) for use in
different languages or typesetting traditions.

Best wishes


chris