> >Show me a widely used font which contains both U+03C6 and U+03D5.
>
> That was not the issue. The issue is when font wanted to add 03D5
> that they would not just put the opposite glyph into 03D5. Or just
> end up having a duplicate glyph. Fonts that have 03D5 by their
> nature should be intended for use with technical publishing and
> therefore are not (as) free in their choice of glyph for 03C6 as
> pure text fonts are.
OK, let me formulate my provocative question differently:
Show me a widely used (technical) font which contains the Unicode
3.0 shape of U+03D5.
I really wonder that no major font company has ever addressed this
problem publicly (at least this issue has never been discussed on the
OpenType list AFAIK) -- Unicode 3.0 is out since a long time...
Here my proposal to partly fix the problem. It won't help for
pre-Unicode 3.0 documents but it should enable software to use older
fonts which use Adobe Glyph Names with recent Unicode.
. Adobe should fix the mapping in AGL's `glyphlist.txt' since the
AGL identifies glyph shapes. Thus `phi' is the stroked glyph and
`phi1' the curly version. I'm referring to Adobe's `Symbol' font
version 001.007 -- most PS printers have this font built in.
phi 03D5
phi1 03C6
. The annotation to U+03D5 should not refer to the AGL entity `phi1'
but to `phi' (and something similar should be added to U+03C6).
. If both PS glyph names `phi' and `phi1' are available in a single
font, the software should rely on them instead of mapping Unicode
code points to glyph indices directly. Otherwise proceed as
usual.
. In the OpenType specification, the `post' table should get a new
version 3.1 (or something similar) to indicate the use of Unicode
3.0 and newer for glyph shapes. Alternatively, a new GSUB
feature called `uni3' which flips the two glyph shapes could be
added. [This is a weak point. Any better ideas how to mark an
OpenType font to be compliant with Unicode 3.0?]
Given the fact (as shown in another mail from Raymond) that most fonts
haven't been updated to the Unicode 3.0 glyph shapes of phi I can
imagine that my proposal greatly reduces the number of incorrect
displays of U+03C6 and U+03D5.
Werner
--
Linux-UTF8: i18n of Linux on all levels
Archive: http://mail.nl.linux.org/linux-utf8/
