On 25/1/14 03:14, Dohyun Kim wrote:
Even after removing syllable decomposition table in ccmp and
recompositon table in liga, jieubsida fonts will still have an issue
regarding the case 4, that is multiple jamos to be composed to single
jamo. If current hangul shaper could be modified so that it would not
touch upon this sort of jamo sequence, then everybody would be happy.
In other words, I propose that hangul shaper should not compose jamo
sequence as follows to precomposed syllable even though it is
composable:
<L, L, V> should not be compsed to <L, LV> but leave it as untouched.
<L, V, V> should not be composed to <LV, V> but leave it as untouched.
<L, L, V, T> should not be composed to <L, LVT> but leave it
as untouched.
<L, V, T, T> should not be composed to <LVT, T> but leave it
as untouched.
Can this request be easily imported to current hangul shaper? If
positive, then the remaining process could be done with opentype gsub
features such as ccmp, calt, or other "global" features, even if *jmo
feaures are not applicable. If the answer is negative, I don't care
that much however, as these jamo sequences are not valid under the
domestic standard KS X 1026-1.
We could do this, although I'm a bit reluctant to do so, as discussed in
my response to Matthew.
In addition to my main concern - that making non-canonically-equivalent
sequences behave as though they were identical is actually unhelpful,
because it creates hidden incompatibilities with other text and systems
- there'd probably be a (slight) performance impact on all Korean text
because of the additional checking of context.
Then there'd also be the question of whether a sequence like <L, L, LVT>
should actually be decomposed to <L, L, L, V, T>, in case the three L
jamos want to compose themselves into a single complex L. If we don't do
this, cases that *are* canonically equivalent might no longer render the
same.
I really don't think we want to start down that path.
JK
_______________________________________________
HarfBuzz mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/harfbuzz
