Peter,
>
> >Fujitsu implemented it in the font rendering engine.
>
> But they were not dealing with complex-script support. That makes a huge
> difference. If there were not any glyph transformations between the cmap
> lookup and rasterisation, your idea indeed would not require
> major changes
> to the system. But we must assume a glyph transformation engine that lies
> between those two points.
>
I am presuming that you can use the power of the transformations to make the
job easier.
>
>
> > Many fonts such as
> >TrueType are encoded in Unicode even if the text is in code page. Thus
> Euro
> >would always be x'0020AC' even if you are running a Widows code page
> with the
> >Euro mod.
>
> Yes, but if the software is using an old codepage, e.g. the original
> codepage 1252, then 0x80 will be getting mapped in the codepage
> to U+0080,
> and that is what will be retrieved from the font. My point still stands.
>
If the font cmap is Unicode encoded, then the print system must convert 1252
to Unicode first.
>
> >Using glyph positioning you can with the worst variation limit the
> number of
> >glyphs to 87.
>
> OK, this is an interesting idea that could work: at cmap-lookup time, the
> layout engine identifies characters that are not supported in the
> cmap and
> instead inserts a sequence of four or six glyphs (or perhaps more with
> your trick for indicating planes -- I don't recall the details)
> while also
> setting whatever features are needed to get GPOS lookups (or
> equivalent in
> systems other than OpenType) to happen.
> Why 87 glyphs? I.e. why so many? Couldn't it be done with 16 plus
> what you
> need for indicating planes?
>
87 is the worse case implementation using very a simple glyph positioning
algorithm.
____________________________________________________________________________
The point of this proposal is not to provide a solution for all cases but
rather a standard way to provide a high end answer to the problem of missing
glyph handling. This would be a solution that would provide more
information than a simple single missing character glyph. I suspect that it
would be best to have a font expert work on a prototype solution to
determine how difficult it would be to implement.
One might even be able to use something like font linking to add a font of
last resort that would handle this condition. However I am not sure that
this solution might not need font engine or open type modifications.
Carl
This archive was generated by hypermail 2.1.2 : Tue Aug 06 2002 - 12:21:32 EDT