Re: [RFC] Making default spinbutton values more consistent with the defaults for HTML5's input type="number"

On 2016-01-17 7:53 PM, James Teh wrote:
> Hi Joanie,
> On 15/01/2016 5:16 AM, Joanmarie Diggs wrote:
>> One possibility which was suggested is that ARIA use "-1" as the value
>> of the attributes which are missing or not a number.
> My major concern with this is that -1 could be a real value. Is there 
> anything that states that the values can't be negative? What if 
> someone specified a minimum of (literally) -1 and a maximum of 1? We 
> wouldn't want that minimum to be interpreted as "unknown". Am I 
> missing something here?

I should have picked up on this earlier:  aria-valuenow, aria-valuemin, 
and aria-valuemax are all of type "decimal number", meaning they can be 
positive, negative, or zero.  Using a special case of -1 to denote 
"unknown" or "open-ended" is a non-starter. The idea of using -1 came 
from its use with aria-setsize to indicate an unknown size.  But, there, 
a set size cannot be negative.

So, using "-1" as a special value for aria-value* is not going to work.

>> My questions are:
>> 1. Will that pose any problems for any ATs or platform APIs?
> Probably nothing that can't be addressed in the platform mapping 
> guidelines. In IAccessible2, IAccessibleValue's methods should return 
> an empty value (not -1) if there is nothing to return. So, the mapping 
> to IA2 would need to account for this. NVDA actually only uses 
> aria-valuetext at the moment, so any implementation we do could 
> account for this change.
>> 2. For aria-valuenow, would it make more sense to set the value to the
>>     empty string if it is missing or not a number?
> IMO, this makes more sense for all three.

Makes sense.

>> 3. What do people think about nixing what we have (just for spinbutton
>>     values) and replacing it with text boiling down to user agents doing
>>     whatever they'd do given an HTML5 input type="number" with missing
>>     or non-numeric values?
> This would certainly make things easier from a consistency point of 
> view; user agents then know they just handle it the same way (probably 
> with the same code). It seems potentially odd to refer to another spec 
> like this (does it make ARIA harder to apply to stuff other than 
> HTMl/does this matter?), but I'm sure there are others better 
> qualified to comment on that.
> Thanks,
> Jamie
> -- 
> James Teh
> Executive Director, NV Access Limited
> Ph +61 7 3149 3306
> Facebook:
> Twitter: @NVAccess
> SIP:


'Array(16).join("wat" - 1) + " Batman!"'
            - G. Bernhardt -

Received on Monday, 18 January 2016 19:52:13 UTC