Re: [selectors-4] :blank pseudo-class

On 8/23/2015 7:47 PM, Brad Kemper wrote:
>
>> On Aug 23, 2015, at 12:52 AM, Patrick Dark <www-style.at.w3.org@patrick.dark.name <mailto:www-style.at.w3.org@patrick.dark.name>> wrote:
>>
>> On 8/19/2015 5:17 PM, fantasai wrote:
>>> http://www.useragentman.com/blog/2012/11/05/css3-pseudo-classes-and-html5-forms-their-limitations-and-a-possible-solution/
>>>
>>> The current :blank pseudo-class should probably be renamed
>>> and :blank used to indicate blank (omitted) form fields.
>>
>> I think :empty-or-white-space would be an intuitive substitute for :blank. The name is verbose, but I don't think it matters since this pseudo-class doesn't seem like something that would see frequent use.
>>
>> I'm not convinced that an empty-or-white-space pseudo-class should exist at all though. Such a pseudo-class would seem to encourage poor coding practices like the one shown in the spec: a paragraph element containing nothing aside from a line-break. Another poor-form construct that I've often seen involves putting non-breaking spaces in elements that could have otherwise been selected via the empty pseudo-class. Some such constructs like <td>&nbsp;</td> do have a use-case that might be worthy of accommodation by CSS,  but I think that would be better done by adding a line-height property-based unit so that one can use code like <td/> in tandem with code like td:empty { min-height: 1lh; }.
>>
>> I don't think blank is a good name for a pseudo-class referring to blank form fields though. I'd propose one of the following alternatives:
>>
>>   * :blank-text-field, which is more intuitive.
>>   * :no-value, which adds utility that extends beyond text fields and makes the :indeterminate pseudo-class redundant.
>>   * :unmodified-by-user, which allows selection of text fields with default values that aren't an empty string and is more useful for elements like the HTML meter and select elements.
>>
>
> I’m of the opinion that having really long pseudo-classes is not a good idea. It is better to have something that is less burdensome to type and quicker to read, even if is a little abstract.
>
> While ‘:empty’ is better than nothing, I think that the new property (:empty-or-white-space, or whatever we end up calling it) will replace most reasons for using ‘:empty’ on the Web. That’s my gut feeling, anyway. So we don’t really need to be too explicit in the name for how it is different from the soon-to-be-nearly-obsolete ‘:empty’. Using a word that is a near-synonym would be better than making the pseudo=class several words long. If we can find a short word that conveys “empty of visible content”, or “empty of characters other than those that would normally collapse down to a single space”, that should be what we shoot for.

I think I'm leaning toward a conclusion that this pseudo-class is unnecessary (which would make naming a moot point). The single use-case presented in the spec isn't something that should be encouraged, and your replacement hypothesis gives me the impression that the :blankproposal is just a less restricted form of :empty meant to accommodate poor code where :emptywould otherwise work. Or perhaps this pseudo-class proposal is meant to standardize the corresponding Mozilla property (mentioned in the spec) which itself isn't justified.

> I’m personally OK with ‘:space”, even if what it matches might contains no space characters or any visible space in the markup. If you think of outer space as the metaphor, where it’s mostly a vacuum, but where there may or may not be a few molecules floating around in any particular region, then I think that is apropos enough. On the other hand, metaphor isn’t as directly applicable as existing terms like ‘white space’, so, I don’t know.

The term "space" isn't consistent with the existing white-space property though, and even with a name like white-space, the terminology doesn't make sense because the proposed pseudo-class will select empty, replaced elements—often with visual output and without a DOM tree containing white space—as well as elements that aren't drawn at all—many empty elements and replaced elements like audio.

If this proposal must go forward despite the above, :no-contentseems like a decent alternative to :empty-or-white-space since white space is arguably not content. The name would make more sense though if the pseudo-class was redesigned to exclude replaced elements. It doesn't make a whole lot of sense to have a pseudo-class with a name reflecting nothingness or space select an element like img, for example, when clearly there is something there and there isn't necessarily any white space.