Editing a bounce should break connections using that bounce

[themaxhero]

While discussing with @kress95 we had some thoughts about its implementation:

How should we check if a bounce is being used?

There are three options for that:

1. Reference counting, anytime something uses a bounce it would do a dispatch that increments the reference count (and the reverse is also true):

• This method is simple in theory but complex in practice, forgetting to dispatch would lead to incorrect behaviour.
• Dispatches should happen after events, so tabs can keep correct reference counts.
• I'm not a fan of this method, it's simple but error prone.
• This method is the fastest to run.

2. Checking tunnels. but they aren't properly implemented yet:

• This method is complex and slow but fast enough for single check.
• The code is complex depending on Tunnels implementation, if tunnels are owned by servers, the function would have to fold every server.

3. A request would tell us if a bounce is being used or not:

• This method is the slower to run but faster to code.
• Using this method is a PITA, it requires two messages to perform one thing: one to ask to edit a Bounce, another to check the request response.

Should we break all active channels that use this bounce?

We're assuming that editing a bounce kills every active connection using it, but does it apply to channels? A channel using that bounce should also be killed?

---

@pedrohlc @renatomassaro

[chrisfls]

I'm leaning toward the 2nd method, but my opinion might change depending on the outcome of the later problem (or even other not discussed yet problem).