Allow contributions of custom CSS properties to CSS extension

[octref]

The motivation is that today, css Language Service would suggest that you have invalid properties.

microsoft/vscode-css-languageservice#68

Although by adopting MDN data we should solve the problem, however if new properties come up CSS LS can still fall behind.

We should provide a config [css/less/scss].validate.acceptedProperties or something similar that stops CSS LS from complaining about "unknown property".

[octref]

@aeschli Actually today we already have a config css.lint.unknownProperties, you can set it to ignore to suppress this error. Can you link to original issues that describe the issue?

[usernamehw]

Isn't this setting ignores all properties? For instance, this will pass:

body {
    teapot: yes;
}

[octref]

@usernamehw My worry is that we are offering too much surface area for linting settings, when all we offer is a subset of https://github.com/CSSLint/csslint and now we are forking with some more options...

When we have MDN data for properties it'll actually solve all three existing "unknown properties" issues:

• overscroll-behavior: [css] Unknown property: 'overscroll-behavior' #46866
• font-display: [css] Add support for 'font-display' CSS property #46423
• backdrop-filter: [css] add backdrop-filter (experimental) #27204

[usernamehw]

Then this issue is not really needed.

[octref]

Yea looking for @aeschli’s input

[aeschli]

@octref yes, you can use "less.lint.unknownProperties" to turn off all validation warnings.
But the idea is to just be able to pass a list accepted properties, so you can still get validation for all others.
You still want to get a warning when there's a typo in the name. But you want to get rid of that error on 'appearance' that is not yet a standardized property but well understood by your cssnext post-processor. Given that post-processors become more and more common, custom properties getting more used to.

[octref]

@aeschli The MDN data contains experimental properties such as appearance. CSSNext only implements what's already in Recommendation Candidate, by which point MDN would already have a data point on it.

I suggest that we pull in the MDN data, see if that addresses people's complaint

Given that post-processors become more and more common, custom properties getting more used to.

If the need arises we can always add that. But I think if someone is using a property not known on MDN (as compared to now known on our outdated css-schema), it's fair we output a warning.

[shockthetoast]

Would it be possible to fix the type in the title of this issue? As it is, it doesn't show up in a search for "unknown property". It would help those of us searching for this exact issue know it is being worked on. Thanks!

[octref]

@shockthetoast Done, thanks!