Default template engine should throw for literal property binding.

[theengineear]

I think I’ve only ever seen a literal property binding by mistake, so I’m tempted to keep this strict. However, I wanted to farm for dissent a little bit. Currently…

NOT OK

html`<div .foo="bar"></div>`;

OK

html`<div foo="bar"></div>`;

or…

html`<div foo="${'bar'}"></div>`;

^^ But this last one is considered an anti-pattern and should not be used.

[theengineear]

IMO, we should not allow .foo="bar" — this binding can be expressed within the html syntax. And, if it’s really needed (e.g., the custom element doesn’t do attribute syncing), you can use .foo="${'bar'}" as a last resort.

[theengineear]

🤔 — I actually think we should do what we’re currently doing… just totally ignore that. You can do this today:

document.body.innerHTML = `
  <div .foo="bar">
    I have an attribute named dot-foo with a value bar!
  </div>
`;

I want to try very hard not to mess with default behavior… I wonder if that means we ought to actually not fail in the case of:

html`<div .foo="bar" .another="${whatever}"></div>`;

… not totally sure what the “best” choice here is!

[theengineear]

While the spec allows for attributes with a preceding . — we are already going to further-restrict attribute names. Therefore — this ought to be naturally rejected by the following logic:

1. Properties must have binding (i.e., cannot be literals).
2. Therefore something like .foo="bar" is considered an attribute.
3. Attribute names may not start with a ..
4. Therefore — we will throw for templates like this.