Polymer encouraged incorrect document.createElement() usage

[annevk]

See https://bugzilla.mozilla.org/show_bug.cgi?id=1294100 and references from there. It seems like we need to support document.createElement(string, string). Not entirely sure if ignoring the argument or treating it as id is better.

Sigh.

[domenic]

@dominiccooney, have we seen fallout from this in Blink?

Ugh.

I'd definitely prefer ignoring the argument; we should not opt in old code to the new semantics...

[annevk]

@smaug---- was saying the same.

[WebReflection]

My 2cents: document-register-element now supports 3 document.createElement overloads: (string), (string, stringIs), (string, object{is:string})

After all I don't see why the is case should fail in polyfilled browsers, sent either as string, like it was already in v0 specs, or object with an is property.

[domenic]

The reason it should fail (ideally) or do nothing silently (as is probably what we'll have to spec) is that the semantics have changed. It is the same reason we had to rename all the APIs. We cannot have old v0 code transparently upgraded to v1 semantics, since they are different.

[WebReflection]

So, if I understand correctly, passing an object should be V1 only and passing a string V0 only?

[domenic]

Yes (in browsers that implement v0, i.e. Chrome only).

[WebReflection]

Thanks, mine right now is a polyfill for v0 but I'm planning to bring in V1 too.

Although I thought WebKit didn't agree about the is property, did they change their mind? If the V1 is final and agreed by all vendors then it's time to polyfill it.

[domenic]

WebKit does not agree with customized built-in elements, but they are included in the standard since we have multiple vendors interested in implementing. If it turns out that for some reason that doesn't happen, they will be removed.

Nothing in spec land is ever final, however. It's a judgment call when you want to start implementing something. All major web browsers already have.

[WebReflection]

I understand nothing in spec is never final but V0 somehow is, and V1 hopefully is as well.
Versioning is a common "nothing ever final" approach 'cause vendors/implementors need to actually implement something so ... is V1 still heavily discussed or something worth polyfill?

I'm just trying to have an ETA for my poly users and myself too. Thanks

[domenic]

v0 is abandoned, not final. v1 is still being heavily discussed as it is being implemented.

[rniwa]

Like we've stated multiple times in the past, we're not going to implement custom builtin elements in WebKit but all other aspects of custom elements should be interoperable across browsers.

[annevk]

I guess that would be another way for Fx to dodge this. Just not implement custom builtins…