[whatwg] Week Strings

[Ian Hickson]

On Fri, 19 Jun 2009, Smylers wrote:
>
> For <input type=week> elements the spec requires:
> 
>   The value attribute, if specified, must have a value that is a valid
>   week string.
> 
>     -- http://www.whatwg.org/html5#week-state
> 
> But the spec's HTML source contains this comment immediately afterwards:
> 
>   <!-- ok to set out-of-range value, we never know when we might have to
>   represent bogus input -->
>
> Does that comment mean that the above requirement will be changed to
> something along the lines of "must have a value that is a syntactically
> valid week string but may represent a week that doesn't actually exist"?

No, it just means that there's no requirement that the value="" be within 
the range given by min="" and max="". It's a reminder to myself in case I 
notice there's no such requirement one day and go and add one thinking it 
was a mistake.


> That is, the author can seed a browser's week-picker control to a value 
> which the browser must not submit back to the server?

So long as it is a valid week string, yes.


> In general determining that something is a valid week string requires 
> knowing which day of the week the year in question begins on.  For 
> example "2009-W53" is a valid week string (because 2009 began on a 
> Thursday) but "2010-W53" isn't (because 2010 will begin on a Friday). 
> Browsers will need to do this to know whether they can submit a week 
> value.

Yup.


> The spec doesn't appear to provide an algorithm for determining which 
> day of the week a year begins on (however I am not a browser developer; 
> possibly this is sufficiently straightforward that those who are don't 
> need it spelling out).

As far as I can tell, it is well-defined in the spec.


> Currently Validator.nu accepts this:
> 
>   <input type=week value=2010-W53>
> 
> but not this:
> 
>   <input type=week value=2010-W54>
> 
> If out-of-range week values are to be permitted in <input> elements then 
> a validator should permit both of them.  Conversely if they aren't 
> permitted then it should accept neither of them (and therefore have to 
> implement a 'which day is January 1st' algorithm, which I'm guessing it 
> currently doesn't).

Please report such bugs straight to Henri. :-)

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