[TheSwordsman]

Telemetry is necessary to be able to observe the system, and look for adverse impact. You should be more thoughtful to the people supporting the tools you use, because without telemetry they do a bad job keeping it working for you.

[TeMPOraL]

If anything, the deterioration of quality in most modern software is a proof that telemetry makes people do bad job at keeping software working for its users.

[scrollaway]

Saying "the deterioration of quality in most modern software" is such a cop-out. There's no universal agreement upon any "general" deterioration, and I'm not sure you're keeping track of the "deteriorating" software that has telemetry vs. the one that doesn't. I personally find that a lot of software I use daily does improve over time, especially web software.

You want a counter-example? Reddit has very little telemetry and quite famously barely looks at the data it does gather. You want to talk about deterioration, how's that for some severe rot.

[TeMPOraL]

> Saying "the deterioration of quality in most modern software" is such a cop-out.

Fair. It's just my opinion. Though I'm not the only one expressing it. You've probably heard the phrase "optimizing for lowest common denominator", or as 'dredmorbius calls it, "the tyranny of the minimum viable user".

> I personally find that a lot of software I use daily does improve over time, especially web software.

I find the reverse. GMail and Dropbox being prominent examples.

> Reddit has very little telemetry and quite famously barely looks at the data it does gather.

Huh. That's not what I expected. I see Reddit as poster child of making the UX worse and worse, driven by advertising goals - something that generally does correlate strongly with running telemetry. I'm confused about them now.

[scrollaway]

> GMail and Dropbox being prominent examples.

Dropbox I'd agree with, gmail I actually much prefer the current UI to the old one.

And indeed web services do tend to optimize for the "lowest common denominator", or more generally for the "majority of users". Which does tend to fuck over power-users. But it also means for most people telemetry works out.

[TeMPOraL]

I disagree that this works out well, because - perhaps unlike the data-driven companies - I don't believe the measure of a good program is the number of registered users. I believe it's just a half of the equation, and the actual equation looks more like (number of users * average utility for user)[0]. Whenever you dumb down your application by removing useful features or sacrificing ergonomics for looks, you're trading average utility for adoption. The software is more appealing to more people, but less useful to them[1].

It does fuck over power-users, but it also fucks over regular users. Not only doing tasks takes longer than it could (or than it took in previous generations of equivalent software), it often precludes them from becoming power users. Because a "power user" of a specific suite of software is something a person becomes over time and repeated exposure. Which includes essentially everyone doing a full-time job in front of computers. I believe dumbed down software is causing a huge hidden economic loss in reduced efficiency of office workers. Not to mention their misery.

(A good example here would be POS systems. If you've ever seen a DOS based one, you'll know it's an order of magnitude more efficient to use than the current breed of browser-based ones. The old-school UI was clean, ergonomic, consistent and fully keyboard-operated, allowing to do most tasks without even looking at the screen for most of the time. There was a relevant thread on HN recently[2].)

--

[0] - actually, I think it's more like: $$ \sum_{user \in users} utility_{user} $$ (https://latex.codecogs.com/gif.latex?%5Csum_%7Buser%20%5Cin%...).

[1] - by "useful" I mean, what tasks it lets users accomplish and how efficiently.

[2] - https://news.ycombinator.com/item?id=21045935

[hsiehwnb]

No, it is not. We have been selling software for decades without telemetry and it worked just fine.

I am more than willing to help GitLab, but telemetry in a VCS is simply a red flag (even a legal impediment in many cases).