arsd v10 tagged

Posted 2021-06-14

Of the 44 things on my wish list, I only actually implemented 7, but the rest can wait - I was able to minimize the breaking changes while maximizing the future potential impact in minigui.

Core D Development Statistics

In the community

Community announcements

See more at the announce forum.

What Adam is working on

I tend to release in bursts. I queue up changes, do a release tag, then realize there's a bunch more I want, that I want to include, so I do a bunch of rapid-series tags. Ugh. I'm probably going to tag 10.1 soon (likely Wednesday) with a variety of additional bug fixes.

A few notable things pending release:

  • terminal.d embedded emulator lets you choose to swap ctrl+c (user interruption request) and ctrl+shift+c (copy to clipboard) in the configuration struct.
  • minigui now more consistently follows Windows' lead on Linux re the ampersand character in widget labels. (Windows uses it as a shortcut key indicator, before minigui did not do this on linux, leading to a visual inconsistency. Now it does it on both and the documentation for affected widgets is updated.)
  • minigui now has a flex-basis kind of thing so widgets can specify their initial size to go along with the stretchiness and other constraints. At some point I should probably rewrite my whole implementation of the algorithm but not today, just another hacky addition for now.
  • Updated to silence deprecations on newest dmd.

Adam's Rant

I found myself wishing the work application had some kind of telemetry today, to tell me if anyone actually uses various key shortcuts. But the truth is, even if the telemetry was there, it wouldn't be that helpful.

For example, I'd like to know if the users tend toward ctrl+c to cancel or to copy to clipboard. I could perhaps piece this together - if a selection was made, then ctrl+c pressed, the user probably intended to copy it... but I can't actually know that for sure.

That's the big problem I have with these telemetry things: they give the illusion of knowing a lot more than you actually do. Since the information is incomplete to begin with and can pretty easily be blocked, it is hard to actually rely on it for real action.

There's math that statisticians can do to give you some idea of how likely these situations are, so help you un-bias the results.... but they're still virtually guaranteed to be biased results, yet so many people treat them as authoritative.

People know not to trust user surveys too much, since everyone knows plenty of people don't fill them out, or fill them out in error (if you asked me to list all the keyboard shortcuts I use, I'm not even sure I can. But I'll sure notice it if one I did use suddenly stopped working. My fingers just know them better than my brain.). So it is useful info, but not conclusive.

But with the telemetry, it is easy to forget those similar lessons. Then you act on it... and annoy tons of people.

I'm almost of the opinion that telemetry is outright harmful because of this. If it is used with an appropriate confidence level, it can be useful information to consider... but it is so often misused that I'd rather just not have it at all.