Re: Defined Private Use was: SSP default ignorable characters

From: Philippe Verdy (
Date: Sun May 02 2004 - 16:23:03 CDT

From: "Doug Ewell" <>
> Not all font formats, not even all "smart" font formats, can contain all
> of the property information for every character the font supports.
> OpenType/Uniscribe was mentioned as an example where the rendering
> engine does work that would be done by the font in other systems. The
> division of labor between font and engine isn't the same across systems.
> And even if you can tell the font about the directionality and
> default-ignorability of your characters, there are still issues like
> line breaking and mirroring (and maybe others, or maybe those are bad
> examples) that have to be handled outside the font anyway.
> Putting all the property information inside the font forces the user to
> use *only that font* for his PUA needs. There might be a choice of
> fonts that support a particular PUA usage (such as for Klingon -- Mark
> Shoulson, is this true?) and it would not make sense to require all of
> these fonts to be updated to include property information (if that is
> even possible). Better to store the property information separately and
> make it work for any old font the user chooses.
> Storing the custom properties in the font doesn't really provide any
> assurance that they haven't been altered.

Still it seems legitimate to mark the font explicitly with the private
convention it is supposed to support. For example a font containing glyphs
mapped to PUAs assigned in the 2003 version of the ConScript PUA registry could
be marked as such by including a trace of this private usage agreement. Without
it, nothing will indicate explicitly in the font for which characters its
internal PUA bindings were intented.

A PUA signature URI (URL, tag URN, or UUID) added to that font could explicitly
contain the trace of this agreement. I know that TrueType fonts for example
contain language-specific features, marked with a internal language tag. If such
PUA signature could be added as an optional attribute of a language tag, then
the mechanism to embed this signature would be identical between:
"participating" fonts, encoded plain text documents (using Unicode Language
tags), XML/SGML/HTML documents (using language attributes of elements), and PUA
definition XML files.

May be AAT and Graphite or other fonts formats have a way to embed a standard
language tag (by standard I mean a code which would work according to the RFC
3066 scheme or its successor, where optional attributes are encodable).

> (Of course, all is not perfect
> here -- we need to be careful to use U+100001 for our PUA examples, not
> U+10001, which is an assigned Linear B character. Count your zeros
> carefully!)

I'm sorry to see that I didn't note the number of zeroes when replying to PUA
usage examples.
It's not important here, it was just an unnoticed typo that still makes sense as
this was just a theorical example.

This archive was generated by hypermail 2.1.5 : Fri May 07 2004 - 18:45:25 CDT