> there should be a way to opt-out of telemetry
There should be a way to opt-in. Anything else is egregious.
There should be a way to opt-in. Anything else is egregious.
Software installed on my machine is in a privileged position. It has more access to my personal files than my close friends and relatives. I'd rather that software respected some simple boundaries.
This is not to say that Telemetry should be opt-out, just that you place other justifications ahead of the usefulness of Telemetry (which is fine).
However, quantitative metrics like telemetry are used to answer different kinds of questions.
> How big peoples projects are on average doesn’t matter.
To give concrete examples, at my current company we do not collect telemetry and have regularly encountered problems where some assumption about customer usage (the number of items in the largest project, the size of the largest document, the smallest browser window, the oldest browser version, the usage of a deprecated feature) turns out to be incorrect, sometimes wildly so, and we only find out after a customer finally complains that they've been struggling with something for months.
If we had telemetry we could have known about these issues well ahead of time and in many cases avoided the problem in the first place. We still choose not to have telemetry for many of the reasons presented in other comments, but we should at least be honest that it can provide immense value to the product team.
You're right. Opt-in results in a survey population that is self-selected, and therefore largely meaningless. But compulsory participation in surveys is deeply objectionable; just because it's easy doesn't make it OK.
So you can't have representative telemetry that is also optional (for the user). So firms should stop trying to use telemetry, unless they're OK with telemetry data from a narrow pool of users, e.g. users of Beta versions of software.
I think a very pronounced opt-out is fine, because it means anyone who is remotely against telemetry can leave, but you still get data from those who genuinely don’t care. Opt-in means you only take data from those who are explicitly for telemetry, which is a very small and not representative number.
Opt-out does have to be visible and pronounced though, and not have any drawbacks. “opt-out” through 30 secret menus which are only known to those who read internet forums doesn’t count, because you’re still secretly collecting telemetry from those who aren’t aware; and wording like “opt-in by default” is flat-out misleading (it’s like saying “made by metal artisans” when the “metal artisans” are literally robots). I’m pretty sure it’s also not necessary: even if there’s a big checkbox “enable telemetry” which is on-by-default, or in this case a flag `telemetry: false` which is very well-documented, the majorly of people choose the path of least resistance and will keep it enabled.
Software is best when it's designed and built by an expert with an opinion. You can look at things like Linux or vim as an example.
All software - including Linux and vim - makes some choices that benefit the developer over the user. Any choice of technology or software architecture with consideration of developer ergonomics is doing this. Every app that ever shipped with a runtime abstraction (classes, dependency injection, testable code patterns, debug logging, etc) chose to make the software fractionally worse for the user, to help the developers to actually write the software.
There is a meaningful and debatable tradeoff in all such decisions, and it's not clear cut. It is often (but not always) sensible for software to include extra code that, yes, must be maintained etc, which exists purely to make the development process of the software more effective. All extra code like this is strictly speaking a bad thing for users when considered in isolation, but overall improvements in development processes (better software, fewer bugs, quicker fixes) mean it's a net benefit for the user in the end, hopefully.
In this case, these developers genuinely do think the telemetry will help them build better software. You may not agree! That's an interesting point to explore, but their opinion is that the maintenance cost is worthwhile for the benefits that this information will provide. They think that the downsides of telemetry will be outweighed by the resulting significant improvements for people using the tool for its intended purpose.
They may well be totally wrong - there's an interesting balance there that is worth discussing, in terms of concrete benefits and real-world costs. Reducing all DX code to "literally no value" though is unhelpful and misses the real discussion.
No, as a user I really don't want my usage data leaving my machine. There is no justification worth that to me.