On 15/10/2004 02:36, Richard Cook wrote:
...Surely what is needed is a rendering system which guarantees to do what is strongly recommended in the Unicode standard: render all canonically equivalent sequences identically. The best way to ensure this is for the rendering engine to normalise the text before displaying it. The normalisation should not necessarily be into a Unicode normalisation form (and maybe should depend on what characters are in the font) but it should fold all canonically equivalent sequences. One way to do this might be first to decompose to NFD then to compose again only precomposed characters which are in the font.
If someone wants to normalize my text into precomposed things, that's all well and good, so long as there's a fallback mechanism for rendering via a font which may have only the decomposed parts.
I understand that Uniscribe is being enhanced to do something like this. Other rendering engines should also do it. The overhead of normalisation should be trivial relative to all of the other tasks required for normalisation.
-- Peter Kirk [EMAIL PROTECTED] (personal) [EMAIL PROTECTED] (work) http://www.qaya.org/
