Reduce the amount of work done to assemble a list of all properties

[domenic]

As a follow-up to #278. #210 is probably related?

lib/generated/allProperties.js currently contains all properties in CSS specs according to the W3C JSON file.

lib/utils/allExtraProperties.js claims to contain other properties which are not in CSS specs but are implemented in browsers. I am unsure this claim is true. Also, many of the properties listed here also exist in allProperties.js, so they are redundant.

Both of the above files are only useful as input to scripts/generatePropertyDefinitions.mjs, which uses @webref/css. The end result is: @webref/css ∩ (allProperties.js ∪ allExtraProperties.js).

So basically, we are generating fewer properties than are available in @webref/css, using a pretty complicated system.

I think it's best to delete lib/generated/allProperties.js and lib/utils/allExtraProperties.js and just use @webref/css. Is there any downside to this?

[asamuzaK]

Up until now, we've (very half-baked) implemented properties with the webkit prefix.

I think there are three options:
a. We will no longer support properties with the webkit prefix, except for -webkit-line-clamp, which is currently on the standardization track.
b. We will continue to support as many of the webkit prefixed properties that we have supported until now.
c. We will leave things as-is.

In case a, it's fine to remove allProperties.js and allExtraProperties.js.
However, this approach may result in regression bug reports from environments that rely on webkit prefixes.
In case b, keep the webkit prefixed ones in allExtraProperties.js, and remove the rest (since most of them are in the standardization process).
In case c, do nothing.

I personally support a.

[domenic]

As I wrote in #278 (comment), I support something close to (a). We should support all standardized webkit prefixes. There are a lot of them. They are all included in @webref/css since @webref/css is created by crawling all existing specs.

Most are aliases of non-prefixed properties and so are pretty simple. However https://compat.spec.whatwg.org/ contains several that are not aliases, e.g. -webkit-text-fill-color. Also -webkit-line-clamp is defined in https://drafts.csswg.org/css-overflow-4/#webkit-line-clamp and is more complicated than an alias. Maybe there are others.

In terms of regressions:

• It is OK to cause regressions by removing a property that is not in any specification/not in @webref/css.
• It is not OK to cause a regression to a property that is in a specification/in @webref/css.
  • This includes either removing the property, or changing its behavior in a way that makes it less spec-compliant. For example, changing from special handler code to generic code would be an unacceptable regression.

My suggestion in the OP here and (2) in #278 (comment) causes regressions of the first type. It causes no regressions of the second type.

[asamuzaK]

I think it's okay to delete lib/generated/allProperties.js and lib/utils/allExtraProperties.js

[asamuzaK]

This could easily be included in #278.
Should we?

[domenic]

I haven't looked at your latest changes in #278 (and I am not sure I will get time to do so today, sorry!), but I am happy for you to include this work there :)

[domenic]

I think this is completely fixed by #278.