Consider making `id` optional.

[mmocny]

Element timing entries returns both an element and an id for its timings.

LCP timing entries also have id, but typically juts "".

Container Timing right now, same as element timing, uses "id assignment" as the opt-in mechanism. But often this just leads to generating random identifiers for many use cases, or name clashes.

Consider an alternative:

• Perhaps any containertiming-nesting strategy should suffice to opt-in for container-timing (nesting implies this is, itself, a container).
• elementtiming="id" could still be used to assign an explicit id
• element is returned to help identify the container, but might be null.
  • We can solve for that case by including a querySelector value (this would be a separate feature request for all element timings, a common request), (or the developer can assign an id)
• This also removes the need to have "default" strategies for nesting

[jasonwilliams]

Please look at https://github.com/bloomberg/container-timing/blob/main/docs/container-definitions.md which should address the issues mentioned here

[mmocny]

We discussed this week various requirements / risks of the use of manifest or observer options using querySelectors (flexible, but feels potentially expensive) vs named attributes on elements (rigid, but clear).

I wonder if this is related: https://github.com/WICG/idrefs

[jdapena]

ID is already optional right now. I will be introducing a new property in the PerformanceContainerTiming entries that is rootElement, that provides a weak reference to the container timing root. This should be enough for queries where the container timing root is not removed (and we have all the tools for referencing that node).

The IDs would only be useful for when the container timing root is gone from the document, as the weak reference would be null.

[mmocny]

Also: If PerformanceObserver observe() references a specific container, then you will already know which scope the element paints are for (even if the element is removed and unnamed).

This isn't true if the stream of container timings is global.