[webkit-dev] Fwd: HTML5 & MathML3 entities

[Alexey Proskuryakov]

17.09.2010, в 11:37, Maciej Stachowiak написал(а):

> rang did not used to be (directly) a CJK character it was in the math symbol block at 2329 but erroneously given a canonical decomposition to the CJK block at 3008. Unicode recognised the error but never change canonical decompositions so they deprecated 2320 and introduced a new character 27E8 that is the same apart from this decomposition.

This certainly sounds like it was a mistake originally, but in practice, this meant that ⟩ and ⟨ were rendered as CJK characters by browsers. U+27E8 looks differently (and is possibly treated differently by screen readers), and it's this change in behavior that I find unfortunate.

> Also, even if Unicode had not deprecated 2329, the W3C "Unicode Normal Form" bans entities using characters that have canonical decompositions as it makes entity expansion and NFC canonicalisation interact badly. Changing the definition is unfortunate but sort of inevitable given the history.


I agree that changing the definition in some way was inevitable (and in fact, WebKit had rang/lang to expand to U+3009/U+3008 for a long time).

- WBR, Alexey Proskuryakov

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-dev/attachments/20100917/c9b6457a/attachment.html>