This is rather a DOM spec thing: https://dom.spec.whatwg.org/#dom-element-setattributens

And as can be seen there, it doesn't really care about the value. There's also nothing requiring that the "value" (string) => "something useful for the engine" should happen synchronously when setAttribute{,NS} is called.

We do drop a note in the console though:

 'Error: <circle> attribute r: Expected length, "foo".'

I'd be happy to try and improve that some more.

On a side-note, if setAttribute were to throw for each value it didn't recognize, the platform would likely be in trouble.

Thank you for your reply!

To be clear, my report is not about the attribute parsing but rather it is
about the "drop a note in the console"  approach.

Developers focus a lot of attention on the browser console. It is their
primary UI to develop web pages. As we are trying to use the Web to create
useful content, we request that the browser also try to be useful in its
output to the console for developers.

Sending "Error" to the console as a logging message is not useful.  We
don't have any mechanism for routing such messages.  We can't log them when
users find them; we can't check for them in unit tests; we can't prevent
them when we know the errors are harmless. There is an established and
effective solution for errors: throw JS exceptions or route the errors
through window.onerror.

Thanks.

Regardless of mechanism this sounds like something that would be "Web exposed", so should probably be ventilated outside the Chromium project first before going anywhere. Maybe https://github.com/DeveloperToolsWG/console-object is a good starting point for discussions?

I opened https://github.com/DeveloperToolsWG/console-object/issues/38

Since it's a spec issue, I'll close it here.
Feel free to open it again when spec will be ready.