[OTSpec] Use of term "Character"

Simon Cozens

John Hudson said:

@Simon Cozens I think Paul Nelson, whom I am pretty sure wrote the <ccmp> feature description, used 'character' as shorthand for 'the default glyph mapped to the character in the cmap table'. The point is that <ccmp> is processed very early in layout, so the input is expected to be default glyph IDs from the cmap table (but possibly output from preceding <locl> feature). But yes, as noted, <ccmp> is a GSUB feature like any other, working entirely in glyph space.

That's got to be considered a bug (and not just the misspelling of "decomposition"). Specifications really should use precise language, especially when the term "character" has a different, well-defined meaning in the context. I suggest writing this out in full, or rewording. (Starting a new thread so @Peter Constable and others can find it.)

John Hudson

I've long maintained that the entirety of the OTL feature registry needs to be reviewed in the interests of accuracy, precision, and consistency. We did this for the <init> <medi> <fina> and <isol> features last year, because they were so desperately in need of revision, but the rest of the feature descriptions should be looked at. I also think we should standardise some additional information regarding each, with regard to default state and feature type with regard to pre-shaping, orthographic unit shaping, typographic features, etc.. 

[With regard to norms for specifications, for a long time in the late 90s and early 2000s it wasn't generally agreed whether the OTL feature descriptions were to be considered part of the OT spec or a kind of informative appendix. I don't think that really became clear until the ISO Open Font Format process included the feature registry.]

Peter Constable

I'll revise the 'ccmp' feature description as follows (changed text shown as bold):

"Function: To minimize the number of glyph alternates, it is sometimes desired to decompose the default glyph for a character into two glyphs. Additionally, it may be preferable to compose default glyphs for two characters into a single glyph for better glyph processing. This feature permits such composition/decompostion. The feature should be processed as the first feature processed, and should be processed only when it is called."

I'm not going to do any major work on feature descriptions at this time, but hopefully this helps on this issue.

David Perry

As long as this is being revised a little, should it take into account base characters with multiple diacritics (common in polytonic Greek and Vietnamese, maybe elsewhere)?  Just say "two or more"?

Peter Constable

Sure:

"Function: To minimize the number of glyph alternates, it is sometimes desirable to decompose the default glyph for a character into two or more glyphs. Additionally, it may be preferable to compose default glyphs for two or more characters into a single glyph for better glyph processing. This feature permits such composition/decompostion. The feature should be processed as the first feature processed, and should be processed only when it is called."

Simon Cozens

decompostion => decomposition

Peter Constable

Old typo; I wasn't checking for such. Thanks.