> It seems like I'm confused by the purpose of this change and that some of my comments
> were/are not accurate.
The purpose of changing the defaults for "close editors automatically" would be entirely for performance and memory reasons.
The purpose of changing the default from "reuse" to "open new" would be to improve the user experience.
The purpose of increasing the default number of open editors would be to make the change less noticeable for users who are accustomed to the current behavior.
We could consider increasing it further if you think it would help -- if the number of tabs before closing is larger than what you'd normally let it grow to before running "close all editors", there would be no change in behavior at all -- it would only affect those users who never run "close all editors"... which is good because those are also the users who are most affected by the performance concerns. Would 30 editors be more palatable?
> I'm really using the order of tabs (open editor in my understanding) as a history
The change we're discussing would only kick in for editors that aren't in the 20 most recent tabs. In my experiments, it only seems to affect editors which have already disappeared from the tabs and are currently hidden in the chevron, not tabs which are currently visible... but I can't state with certainty that it would never affect a visible tab (do we still use MRU tab visibility? If so, we could state this with certainty.)
Is there a reason you wouldn't want to use alt-left-arrow rather than ctrl-pgup to navigate to the previous editor? I think it would address all the use-cases without the performance cost. Also, you'd probably want to use the dropdown below the back button rather than the dropdown in the tabs -- but that seems preferable for your use case since the order below the back button really reflects the navigation history whereas the order in the chevron is alphabetical.
> However, I actually do not care if there is actually an editor "active" for a tab that doesn't have
> focus: if the editor is disposed when I don't use it and re-instantiated when I select the tab, that's fine.
I agree that would be ideal... if we could destroy and recreate the editor entirely without leaving any user-visible footprint. Unfortunately, the best approach we support at the moment does leave some user-visible side-effects. Namely: the closed editors disappear from the ctrl-e menu and the number shown in the chevron is lower.
My argument is that this user-visible side-effect is minimal enough to be acceptable given the considerable performance gains that come with it.
> Ok, I trust you. It's just that I don't feel it at usage. Maybe because I'm just accustomed to it.
Since the entire justification of one of the changes pivots on the presence of these performance issues, asking me to back up that claim with evidence you can inspect is perfectly reasonable.
- Stefan