outdated specification for the OS/2 table

MS Font Validator warns about bits set in the OS/2 (notably the "reserved" bits for Unicode block coverage, which are no longer reserved since long).

E.g. bit 96 of the ulUnicodeRange documented for Buginese: if I set it, the validator incorrectly signals that this is incorrectly set.

Peter Constable (in the Unicode Mailing list) said that these script-coverage bits in the OS/2 table are outdated (not sure that it means deprecated), because they are now all allocated, and we already have more blocks than the 128 that were initially allocated in this table.

May be there's a new extension table in OpenType to specify in a font that we have sufficient coverage of that block.

Anyway, a font that only contains the highly recommanded mappings for NULL, CR, and SPACE but no other Basic letter (or digit, punctuation) should not have bit 0 set (the validator incorrectly suggests setting this bit for the Basic Latin block: it should not count mappings in the whole 128-character block, but only in the U+00021..U+007E subset (94 characters) before determining that bit 0 should be set.

Additionnally it should also suggest that the font maps the NULL, CR and SPACE, independantly of this bit.

Same remark for the next block for Latin 1 : the controls should not be counted, and NBSP should be mapped in all fonts that contain mappings for any script with combining diacritics. So bit 1 should not be suggested if there's no mapping in the space U+00C1..U+00FF.

Why all this ? Authors of non-latin fonts should not be forced to create and map glyphs on the Basic Latin space, because this complicates the use of that font in combination with another font offering a much better coverage (and better typography) of the Latin script.

Life would be better for every one, if fonts were designed separately for separate scripts (or coherent but complete subsets of related scripts such as Latin+Greek+Coptic+Cyrillic). I'm fed up of seeing excellent fonts for other scripts overriding the Basic Latin subset and breaking the coverage of Latin performed in a better font (this complicates the design of multilingual documents, because we need to add font-family styling for each change of script, and in CSS there's no way to specify per-script preferences for separate font families).

NULL, CR, SPACE and NBSP are not script-specific, their mapping will be present in all fonts, this does not mean that those fonts have to cover Latin (don't assume that multilingual documents using a non-Latin script and the Latin script will only use Latin for writing only in English ; other Latin-written languages often need much more, and should benefit from a font with good coverage and typography, developed separately).