Open jgraham opened 9 months ago
Is there an issue/discussion for this PR? I wonder if it's worth making context IDs to be session-specific: I see the value of being able to get a context id in one session and start a different session for a different task (e.g., for tracing or network monitoring) on that specific context id.
My assumption has been that as far as possible sessions should be isolated from each other. If you have multiple tools sharing data, they can probably also share a session id and just operate entirely within the same session. This also matches what we already do for node ids.
I think it was brought up by someone at TPAC that there could be cases where you would want to use a dedicated tool (for example, a network capturing tool) that would create a session to do a specific task in the running instance. At the same it might not be possible to use an existing session because the event subscriptions might not be compatible between tools: e.g., the tool or the main program might not handle the events in the same way.
The Browser Testing and Tools Working Group just discussed Make context/navigable id per session
.
Also make it clear that context ids actually relate to navigables, not browsing contexts. This part is a temporary workaround while the rest of the spec continues to pass in browsing contexts rather than navigables.
Preview | Diff