Mateusz Paprocki

We know, there were certain design limitations that prevented this, but those are currently being worked on, so we should be able to add this feature in future.

A fullscreen tool was added in PR #11915, but the implementation requires some tweaking to make it actually useful.

It looks like an unnecessary limitation. See PR #12296. I made it to display the menu on focus for a consistent behavior.

> If pd.NA moves out of experimental status then we need to figure out how to serialize it properly (...) This is easily possible with the new serializer. In particular,...

Yes. However, one thing to note, is that the existing layout requires all widgets to be wrapped in `WidgetBox` to make them any layoutable. Widgets on their own don't participate...

This now work much better and the layout is respected. However, the widget CSS is still lacking, overflowing the layout allocated slots.

Currently attribution is just a bunch of custom HTML and CSS. There is issue #4284 open for making attribution a proper model and bringing it under layout. With this work...

As an FYI, in PR https://github.com/bokeh/bokeh/pull/12083 (scheduled for bokeh 3.1) I implemented various interactive shapes, including spans, with the intention of having in-canvas UI controls, especially sliders. Interactive spans (movable,...

The simplest non breaking way to resolve this, is to create a revert commit and push again to `master`.

In bokeh we dropped constraint system based layout, because of inadequacy of this approach both from functional and performance perspectives. If the decision was up to me at the time,...