On 22/07/2003 20:49, John Hudson wrote:
Thinking about this whole Hebrew encoding/normalisation problem from
the rendering side -- i.e. in terms of smart font glyph substitution
and mark positioning -- it seems me that *if* a character is to be
inserted between two vowels that visually follow a single consonant,
it would be preferable if this were not a control character. If the
character is something that is actually painted as a glyph, it is
incredibly easy to resolve the rendering problems in the font's
composition/decomposition <ccmp> feature by, for example, ignoring
marks while 'ligating' the consonant + extra character to the
consonant glyph alone. This would happen before the mark positioning
takes place, so would ensure that the presence of the extra character
does not break mark positioning in the way that actual control
characters seem to (and which cannot be be included in <ccmp> lookups
because they are not painted).
Of course, if this problem is thought about from the text search,
sort, etc. perspective, the presence of a non-control character --
even an invisible one -- introduces problems, since it changes the
encoding of the word. As noted previously, users are unlikely to know
that they must insert an invisible character into search strings. My
question is whether this is something that can be addressed in search
engines and other places affected, so that this character could be
filtered out during operations? Or, failing that, if we might here
have a use for a new zero-width character that will be ignored during
search and sort operations --- as control characters like ZWNJ and CGJ
are -- but which will be painted -- as control characters typically
are not?
John Hudson
Tiro Typeworks www.tiro.com
Vancouver, BC [EMAIL PROTECTED]
Couldn't the rendering engine simply treat CGJ as a non-control
character with a blank and zero width glyph? Then it could ligate it in
the way you suggest. As far as I can tell there is no other known use
for CGJ and therefore no need for a rendering engine to process it as a
control character. This behaviour of CGJ can presumably be made
independent of its behaviour in searching and sorting.
--
Peter Kirk
[EMAIL PROTECTED]
http://web.onetel.net.uk/~peterkirk/
