[css-lists-3] Inconsist `counter()` syntax between CSS 2 and 3

[xiaochengh]

Should content: counter(c, none) be a parse-time failure, or a valid value that generates empty text?

In CSS 2.1:

• counter() = counter(name, <list-style-type>?) [1], where
• <list-style-type> = disc | circle | square | decimal | decimal-leading-zero | lower-roman | upper-roman | lower-greek | lower-latin | upper-latin | armenian | georgian | lower-alpha | upper-alpha | none | inherit [2]

So content: counter(c, none) is a valid declaration that doesn't generate any text. Chrome implements this behavior.

In CSS 3:

• counter() = counter( <counter-name>, <counter-style>? ) [3], where
• <counter-style> = <counter-style-name> | <symbols()> [4], where
• <counter-style-name> is a <custom-ident> that is not an ASCII case-insensitive match for "none". [5]

So content: counter(c, none) is invalid at parse time. Firefox implements this behavior.

My proposal: We should not allow none. Allowing none in CSS 2.1 looks like a spec bug to me, since it provides little value and doesn't make too much sense. (And the same syntax also allows inherit, which makes no sense at all).

[1] https://www.w3.org/TR/CSS2/generate.html#counter-styles
[2] https://www.w3.org/TR/CSS2/generate.html#propdef-list-style-type
[3] https://drafts.csswg.org/css-lists-3/#counter-functions
[4] https://drafts.csswg.org/css-counter-styles-3/#typedef-counter-style
[5] https://drafts.csswg.org/css-counter-styles-3/#typedef-counter-style-name

[SebastianZ]

It was already agreed on to make none an invalid value in #4163.

As the definition in CSS Lists 3 supercedes the one in CSS 2 (and CSS 2.2) and already excludes none as well as all CSS-wide keywords, I think nothing needs to be done here.

If something needs to be corrected, it is CSS 2.2 that still says that counter(x, none) yields an empty string instead of making it invalid.

Sebastian

[Loirooriol]

So what's strange here is that #4163 claimed that a none counter style was invalid in CSS2, and wanted to align css-lists-3 with that. But actually none was valid in CSS2 according to https://www.w3.org/TR/CSS2/syndata.html#counter and https://www.w3.org/TR/CSS2/generate.html#counter-styles ?

[fantasai]

@Loirooriol https://www.w3.org/TR/CSS2/generate.html#content

<counter>
Counters may be specified with two different functions: 'counter()' or 'counters()'. [...] The name must not be 'none', 'inherit' or 'initial'. Such a name causes the declaration to be ignored.

Generally-speaking, prose restrictions override grammar leniency.

Agree with @SebastianZ that the example in syndata.html needs to be fixed, though. :/ That's definitely self-inconsistent.

[Loirooriol]

@fantasai Precisely, it says "The name must not be 'none'", given counter(name, style). Here we are talking about the style, not the name.

[Loirooriol]

Also note https://www.w3.org/TR/CSS2/generate.html#counter-styles says

All the styles are allowed, including 'disc', 'circle', 'square', and 'none'.

P:before         { content: counter(p, none) }

[SebastianZ]

@Loirooriol is right. The initial claim in #4163 that counter(name, none) and counters(name, "string", none) are invalid was wrong.

Though I agree with @xiaochengh that neither <counter-name> nor <counter-style-name> should allow none, i.e. align CSS 2.2 with CSS Counter Styles 3 for the same reasons as @fantasai outlined in #4163.

Sebastian