Re: Variation Sequence Problem
From: Mark Davis
Date: 2002-03-28

At the last minute in the release of 3.2, we were notified of a problem with the variation selectors. The UTC agreed after much discussion, that we must restrict the application of VSs to non-decomposables, non-combining marks. As it turns out, we have made a mistake in two cases.

Because it was too late to make a change, we simply added the following note explaining the situation:

Note: Just before publication, an inconsistency was discovered between the above principles and the standardization of the two variant sequences <2278, FE00> and <2279, FE00> because U+2278 and U+2279 are in fact decomposable characters. Those variant sequences denote glyph variants of these mathematical symbols with a vertical line instead of a slanted line as the diacritic to indicate the negation.

The sequence <2278, FE00> is canonically equivalent to <2276, 0338, FE00>, and the sequence <2279, FE00> is canonically equivalent to <2277, 0338, FE00>. So that these equivalent sequences are given equivalent rendering treatment, the use of U+FE00 would have to be interpreted—exceptionally—as defining a variant appearance for the entire sequence.

Because a combining vertical line overlay, U+20D2 COMBINING LONG VERTICAL LINE OVERLAY, is also available in the Standard, an alternate way of explicitly indicating these particular variants already exists. That alternative mechanism is a safer and more stable way to indicate the distinction, as the inherent complications in allowing variation selectors to follow combining marks may require future corrective action to remove the exceptional variant sequences <2278, FE00> and <2279, FE00> from the table.

[See http://www.unicode.org/reports/tr28/#13_7_variation_selectors]

Especially in light of the fact that there is an already-encoded way to express the variation, the UTC should issue a corrigendum removing the variant sequences <2278, FE00> and <2279, FE00> and request that WG2 do the same in a corrigendum to 10646.