From: Philippe Verdy (verdy_p@wanadoo.fr)
Date: Tue Jun 07 2011 - 00:46:42 CDT
Ok, but one week ago (when the message from Mark Davis was posted
here), the same validation tests returned 2 errors (with a red
message), not warnings as of now.
There must have been a recent correction in the validator (partly
corrected, because one additional warning is effectively counted but
apparently discarded/filtered out, or the same warning is counted
twice, possibly after following an alternate execution path for
non-NFC texts).
For lexing/parsing any document using an XML, or HTML, or SGML syntax,
normalization should not be used at all (only parsing of individual
codepoints that may be encoded on multiple bytes, just in order to
support conversion of the encoded document to a stream of Unicode
codepoints). Then for HTML parsing (not XML parsing), only basic
character properties are needed (for letter case).
If normalization occurs, this is much later (it is not even needed for
validating identifiers, or when creating the DOM view of the
document,here in a way compatible with the HTML5 DOM) ; normalization
MAY possibly be applied only when rendering the DOM, or when support
user interactions with the document, such as searching/matching with
user input, in interactive documents/applications).
But plain text rendering should work without having to normalize the
text contained in text elements or attribute values (text renderers
will just perform a best effort work to make it readable, using their
instatalled fonts and supported text features : normalization may help
sometimes, as it suggest strong fallbacks that should be supported by
canonical equivalence, but this equivalence does not imply that
normalization must actually be performed, notably because NO
normalization form will be better than another one, which is
canonically equivalent, including custom forms).
Things to consider with possible normalization : for example matching
anchors in URLs, or perming searches of elements in Javascript, or
validating user input also in Javascript (this is just
application-dependant, for each web site for example)...
2011/6/6 "Martin J. Dürst" <duerst@it.aoyama.ac.jp>:
> On 2011/05/27 0:57, Andreas Prilop wrote:
>>
>> On Wed, 25 May 2011, Mark Davis wrote:
>>
>>> But that needs to be distinguished from saying that
>>> there is something wrong with 20B9 or with NFC.
>>
>> I did not write that there is something wrong with NFC.
>>
>> I complained that HTML5 or validator http://validator.w3.org/
>> *requires* NFC.
>> This might be a bug in the validator and not actually
>> a requirement of HTML5.
>>
>> Validate my test page
>> http://www.user.uni-hannover.de/nhtcapri/temp/yerushalayim.html
>
> Just for the record, it says (all in green):
>
> This document was successfully checked as HTML5!
> Result: Passed,
>
> and then (in yellow) says "3 warning(s)", and shows two actual warnings
> related to NFC. (there seems to be a counting error)
>
> So HTML5 does NOT require NFC, but recommends it. If you know better, use
> something else.
This archive was generated by hypermail 2.1.5 : Tue Jun 07 2011 - 00:49:24 CDT