[whatwg] Steps for finding one or two numbers in a string

[Ian Hickson]

On Fri, 14 Apr 2006, Henri Sivonen wrote:
>
> The part of the spec about numbers at 
> http://whatwg.org/specs/web-apps/current-work/#numbers talks about 
> skipping characters in the class Zs. Also, it talks about characters 
> like U+066A ARABIC PERCENT SIGN, U+FE6A SMALL PERCENT SIGN and U+FF05 
> FULLWIDTH PERCENT SIGN.
> 
> On the other hand, when the spec talks about digits, it talks about 
> "U+0030 DIGIT ZERO to U+0039 DIGIT NINE" and not about character class 
> Nd.

I'd be ok with talking about more numbers if they are indeed potentially 
used in these scenarios and their processing can be clearly defined.


> Referring to character classes instead of particular characters is 
> problematic, because what constitutes a conforming number changes when 
> Unicode introduces ANCIENT BRUTOPIAN QUARTER CUBIT SPACE or SYLDAVIAN 
> VULGAR DIGIT FIVE.

It changes to become more permissible, which seems ok to me. What's the 
problem scenario?


> I think i18n political correctness has no place in attributes. I think 
> they should be ASCII only with the XML notion of whitespace.

I agree and believe the spec already requires this.


> However, the human-readable fallback content of the progress and meter 
> elements is not so clear-cut. Still, I doubt whether it is worthwhile to 
> have an unstable but universal number format for content that is 
> displayed to end users only as fallback.

How is it unstable? Unicode never drops characters, and the characters 
are only added to classes when they don't previously exist.

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'