> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf
> Of Edward H. Trager

>   I find it hard to believe that precomposed letters are the "preferred approach"
>   for encoding.  Isn't this more like the "legacy approach" , not in the true
>   spirit of Unicode in 2004 ?

Actually, the combining overlays are in Unicode only because of legacy -- think about 
it: the origin of U+0338 comes from bksp + "/" used in legacy technologies like 
typewriters and daisywheel printers. You'll note various characters with overlaid bars 
or strokes that have existed in Unicode since version 1.0 but that were not given 
decompositions to sequences with overlays. Evidently, the design principle of dynamic 
composition was never meant to extend to such cases. Even U+0321 and U+0322 fall into 
this camp.

 
>   Is there any reason why I could not create glyphs for "majuscule/miniscule S with
>   diagonal stroke" in an OpenType, AAT, or Graphite font that would then
>   get used as the preferred glyphs when composing text containing the
>   <U+0073, U+0338> and <U+0053, U+0338> sequences? ...

No, there isn't. However, whereas you can generalize the dynamic composition of (say) 
acute with any base in a font implementation very easily using anchor positioning, you 
cannot do so for overlay marks: you would need to design a specific set of precomposed 
glyphs. 

A very good reason not to dynamically compose with U+0321, U+0322, U+0334..U+0338 is 
the ambiguity of the result. Perhaps with 0338 there is less potential for ambiguity, 
but it's especially obvious for 0321 and 0322: e.g. what should < 006E n, 0321 > 
represent? Should it look like 014B or 0272? For the other marks as well, the 
potential for ambiguity is very real.

Whatever reasons may have been behind the initial design decision not to decompose to 
these overlay marks, we can use such reasons as I've mentioned to convince ourselves 
now of the appropriateness of the decision. Convinced or no, however, that decision in 
the early days was made, the precident was established, and UTC has continued to abide 
by it with decisions as recent as the last UTC meeting.


Peter
 
Peter Constable
Globalization Infrastructure and Font Technologies
Microsoft Windows Division


Reply via email to