> In either cases, the "Soft_Dotted" property is probably overkill on
> the existing <ij> or <IJ> ligatures (should should have been better

There is no point in having a soft-dotted property for the capital
letter...

> named "letters" and not "ligatures") for Dutch. Or is this update
> needed to document officially the expected rendering behavior for
> sequences <ij,accute> and <ij,macron>?

Yes. <ij ligature, combining acute> should give a dotless ij digraph
with an acute accent centred over it; <ij ligature, combining double
acute> should give a dotless ij digraph with an acute on top of each
dotless subletter glyph; I'm by now not sure which is the correct one,
but the first one can only be produced this way.  (And the others are
unrelated to the dotless-i and dotless-j, so keep these two out of the
pot.)

> The main interest of the Soft_Dotted property is not to describe the
> rendering for the character, 

Yes, it is.  I should know, the soft-dotted property was my suggestion
in the first place...  And please read the note accompanying the "public
review issue".  Not all of the characters in my initial list was
actually
given the property, however. This is what the current suggestion
tries to correct.  I know, there are Thai and Khmer letters where a
"glyph
appendage below" is removed when there are other things below, like
a vowel or a subjoined consonant; and there is as yet no property for
that...
(But those appendages don't have any similar combining character below
either.)

> but to document how case conversions
> (lowercase, uppercase, titlecase, folded) can be performed safely on

The soft-dotted property is not primarily defined for case mapping,
even though it is used there too.  Case mapping is documented in the
UCD;
for non-same-always-1-1 cases, they are documented in SpecialCasing.txt.
There is no special rule for the ij/IJ combination (even for Dutch)
there; and
it may be unlikely that there will be one.  It's easier to just use the
ij ligature
characters (which do have the expected case mapping already)...

                /kent k


Reply via email to