[whatwg] Submit controls should be barred from constraint validation in favor of a new pseudo-class

[Mounir Lamouri]

On 12/31/2010 02:20 AM, Ian Hickson wrote:
> On Thu, 23 Sep 2010, Mounir Lamouri wrote:
>>
>> It sounds like currently the specifications want explicitly to have all 
>> submit controls being subject for constraint validation [1] which seems 
>> to be a weird idea. Given that only setCustomValidity() can be used on 
>> those controls, the only use case I see is to set the submit controls 
>> invalid when the form is invalid thus having :invalid style applying to 
>> them.
>>
>> If the given use case is correct, I think we should make all submit 
>> controls barred from constraint validation and introduce a new 
>> pseudo-class allowing to style submit controls when their form is 
>> invalid.
> 
> Form controls are part of constraint validation for completeness only, as 
> far as I can tell. Anything that's visible and submitted is a candidate 
> for constraint validation.

Except <keygen> and <object>.
But do we agree that "it's visible and submittable so it should be able
to be invalid" is not a sane rule? At least, we should not include
non-editable but visible submittable controls like submit controls.

>> Current Gecko nightlies (and next beta, beta7) have a 
>> :-moz-submit-invalid pseudo-class. Submit controls can still be marked 
>> invalid but that might change before the final release if there is an 
>> agreement that :-moz-submit-invalid is a better way to fulfill the need.
>>
>> As a side note for web authors on this list, :-moz-submit-invalid 
>> currently has a default style on Firefox but that is very likely going 
>> to change for beta8.
> 
> Do people use this feature much?

It's not in a stable release yet so it's hard to say.
Actually, I was wondering if we couldn't introduce the :invalid
pseudo-class to the form element. This will remove the need of
:-moz-submit-invalid and certainly help for other situations.

--
Mounir