Re: [widgets] test suite, the width/height attribute

On Tue, Dec 15, 2009 at 9:57 AM, Cyril Concolato
<cyril.concolato@enst.fr> wrote:
> Hi Marcos,
>
> Le 14/12/2009 16:49, Marcos Caceres a écrit :
>>
>> On Fri, Dec 4, 2009 at 5:04 PM, Cyril Concolato<cyril.concolato@enst.fr>
>>  wrote:
>>>
>>> Dear Widgets-experts,
>>>
>>> While checking some of the tests, I found some unclear processing with
>>> regards to the width and height attribute of widget element. The spec
>>> says:
>>>
>>> "If the width attribute is used, then let normalized width be the result
>>> of
>>> applying the rule for parsing a non-negative integer to the value of the
>>> attribute. If the normalized width is not in error  and greater than 0,
>>> then
>>> let widget width be the value of normalized width. If the width attribute
>>> is
>>> in error, then the user agent must ignore the attribute."
>>>
>>> It explicitely says "greater than 0" which means that 0 should not be
>>> allowed, but the test suite says for c9.wgt that the result should be 0.
>>
>> Argh. Right.
>>
>>> This seems inconsistent. On top of that, the spec seems to make the
>>> distinction between 'null' (when in error) and '0' (not specified). From
>>> an
>>> implementation point of view, I would prefer two cases:
>>> - specified, not in error, greater than 0, width = the specified value
>>> - in error or not specified, width = null, empty or 0.
>>> Actually, I would prefer 0 since then the attribute can be implemented as
>>> an
>>> integer not as a string.
>>>
>>> What do you think ?
>>
>> Given that a number of UAs have implemented support for getting back
>> the value "0", I think we should just say "greater than or equal to
>> 0".
>>
>> So:
>>
>> <widget width/height="">  = Error. value remains null.
>>
>> <widget width/height="             ">  = Error, value remains null.
>>
>> <widget width/height="abc">  returns 0, value is 0.
>>
>> <widget width/height="100abc">  returns 100, value is 100.
>>
>> <widget width/height="000100abc">  returns 100, value is 100.
>>
>> However, I'm open to just saying return 0 upon error.
>
> That's what I would prefer.
>

Me too, that's what the spec says now.


-- 
Marcos Caceres
http://datadriven.com.au

Received on Tuesday, 15 December 2009 14:56:30 UTC