John Hudson <tiro@...> wrote on 04/02/2003 12:27:08 AM:

> I can't see any advantage to such atomic encoding in a modern text
> processing environment with smart font support. It is easy enough to map
> e.g. b + palatal hook to a ligature using a glyph composition feature.

In a unicoRe thread last fall,

Ken Whistler wrote on 2002-10-15:

Just another example of the line that had to be drawn for
decompositions. Diacritics involving baseform shape alterations --
with the exception of cedillas and ogoneks, which were treated
as honorary non-spacing marks below -- were excluded from the
original set of decompositions published in Unicode 1.1 in 1993,
and later corrected and rationalized in Unicode 2.0, with data
tables distributed...

As far as I know, the same completeness issue does not apply for the
retroflex and palatal hooks -- so for those, use of the preformed
base letters is probably the better recommendation, rather than use
of the non-spacing diacritics together with ligature tables in the fonts.


Michael Everson wrote on 2002-10-16:

All of these things should be encoded as things-without-diacritics,
as unitary characters.

Also, earlier in 2002 in offline discussion Ken wrote,

> I'm also not sure
> how/when 0321 should be used.

Sparingly, I think. It is one of the very small set of actually
attaching diacritics for IPA. As such, it isn't going to work for
rendering on most systems unless you have proper fonts that have
ligature tables to fully-formed glyphs for the combinations.

It is feasible to use it, of course, but generally, the letters with
retroflex or palatal hooks have been separately encoded, anyway.
Cf. U+01AB, not given a decomposition.


My suggestion is just to bring in a proposal [for palatal-hook characters.]
And name them all consistently
as ...WITH PALATAL HOOK, to be consistent with U+01AB...

Given the fact that a precomposed palatal hook precedent exists, and that
all the retroflex hook forms also got encoded -- all with no
explicit decompositions, this way seems most consistent.

In my opinion, neither the retroflex nor palatal hook are
useful as combining marks per se, since they are really
baseform modification diacritics (like the descender for
Cyrillic), and so cannot be used by a renderer for generic
application by rule.

Their usefulness is rather to have a character to indicate
the diacritic per se (as for metadiscussions about IPA
orthography), and as placeholders for such things as the
Unicode Collation Algorithm, which can use them to do the
secondary weighting for characters with those diacritics,
if one chooses.

I was just following the prevailing wisdom. I think the strongest point in
this in the fact that generic application by rule is not possible, unlike
diacritics like acute that can be handled by generic mechanisms.

BTW, Ken, not too long ago we talked about overlaid-tilde typeforms, and
you wanted to use the combining overlay in that case, but it seems to me
that the same argument of inability to have generic application by rule
applies in those cases. But I don't yet have what I'd consider good enough
evidence for proposing a set of tilde-overlaid forms.

- Peter

Peter Constable

Non-Roman Script Initiative, SIL International
7500 W. Camp Wisdom Rd., Dallas, TX 75236, USA
Tel: +1 972 708 7485