[whatwg] HTML 5 clarifications on ValidityState?

[Alex Vincent]

I read this paragraph (from section 4.10.15.3) as self-contradicting:

"The validationMessage  attribute must return the empty string if the
element is not a candidate for constraint validation or if it is one
but it satisfies its constraints; otherwise, it must return a suitably
localized message that the user agent would show the user if this were
the only form with a validity constraint problem. If the element is
suffering from a custom error, then the custom validity error message
should be present in the return value."

Specifically, the last sentence contradicts the rest of the paragraph.
 If there is a custom error, but the element is not a candidate for
constraint validation, should the validationMessage attribute be
empty, or should the custom validity error message be present in the
return value?  You can't have both.

This inconsistency is somewhat worsened when you consider the user may
call input.setCustomValidity("foo") on an element that isn't a
candidate for constraint validation:

<form action="javascript:void()">
<input disabled="true" name="foo" id="foo"/>
</form>

var foo = document.getElementById("foo");
foo.setCustomValidity("foo");
foo.willValidate // returns false, because of the disabled attribute

"The willValidate  attribute must return true if an element is a
candidate for constraint validation, and false otherwise (i.e. false
if any conditions are barring it from constraint validation)."

foo.validity.customError // returns true
foo.validity.valid // returns false
foo.validationMessage // ???

Also, I noticed the readonly attribute for input elements overrides
constraint validation, per section 4.10.4.2.3 - but the readonly
attribute has no meaning for several input types (section 4.10.4, the
big table after the IDL).  For checkboxes, radio buttons, file
uploads, image inputs, submit, reset and ordinary button types for the
input element, readonly has no other effect.

Please advise - I'm currently working on a patch for Gecko code which
implements this.

-- 
"The first step in confirming there is a bug in someone else's work is
confirming there are no bugs in your own."
-- Alexander J. Vincent, June 30, 2001