[cssom-1] Define `CSSRule.cssText` with `readonly`

[cdoublev]

Since 9f8435a.

interface CSSRule { 
  attribute CSSOMString cssText;
  // ...
}

On setting the cssText attribute must do nothing.

Would it be different / a problem if CSSRule.cssText were defined with readonly?

[tabatkins]

Hm, I thought that readonly attributes threw when you tried to set them, but no, the JS bindings definitely just ignore the set, same as what we specify manually. I wonder what the historical reason was for writing it this way?

/cc @dbaron for maybe some insight?

[Loirooriol]

https://webidl.spec.whatwg.org/#dfn-read-only

It is language binding specific whether assignment is simply disallowed by the language, ignored or an exception is thrown.

For ECMAScript, in https://webidl.spec.whatwg.org/#define-the-attributes the setter will be undefined. So per https://tc39.es/ecma262/#sec-putvalue the behavior will be that it throws in strict mode and it silently fails otherwise.

TL;DR: Marking it as readonly would make assignments throw in strict mode.

[dbaron]

This is from well before my time; the IDL definition goes back at least to the first working draft of DOM Level 2 maintained by the DOM Working Group. The text defining that the setter does nothing is substantially newer.

[cdoublev]

Thanks. I did not think to look at the Web IDL spec for readonly.

So it would be different. I assume that making this change this way reduced its impact in terms of backwards compatibility.

I am not requesting a change to readonly. I was mostly interested in why implementing a setter that returns nothing is required instead, so I am closing as this question has been answered.