I'm not beholden to the "proxy" implementation, just the design-properties that it provides: 1) achieves the goal of purging / recovering the memory consumed by the ancestry objects while, 2) doing so safely and transparently, with never any intention of mutilating or dis-membering the ancestry -- because the stub is as good as the original when called upon.
Those are the _design_ properties anyway, if there is some other implementation which could achieve these properties, I would support it.
As for the time-consuming reification, I like Tim's idea of a cancel button on the Progress bar (note that Eliot DID simply use Command+dot) rather than a confirmation prompt because of the headless server situation -- a delay would be painful enough to notice, but not enough to bring the server down like a confirmation prompt sitting there...
I'll at least get it removed from the menu for now.
On Fri, Jan 19, 2018 at 12:47 PM, tim Rowledge tim@rowledge.org wrote:
The problem I have with proxies *as a user* is that when you’re doing something where they help, they're great and when you do something where they surprise you and trigger some time-consuming thing they’re terrible. Finding a middle ground may be difficult.
In a case where something triggers a load of packages - which is almost always going to be pretty time-consuming, even when you have gigabit internet and the package server is in a really good mood - it would be nice to have some opportunity to stop it. In fact I think it would be nice to have a reasonably clean ‘stop this right now!’ button on the progress morph for almost any kind of long-running activity.
tim
tim Rowledge; tim@rowledge.org; http://www.rowledge.org/tim Strange OpCodes: SLTMDL: Shift Left, Test Mask and Dim the Lights