How are companies that aren't software vendors and aren't able to spy on their customers able to do it? Did software companies not have good ways to do this before spying on their users?
1 and 2 are problematic because it's very hard to get representative data from either one. The people who have time for user studies or post on your forums are not representative users.
Only listening to data from 1 & 2 results in the sort of angry posts you frequently see on HN complaining that devs aren't listening to "real users" or have the wrong priorities.
You end up needing data from additional sources, telemetry being one of them.
You do not need it. This is a really weird attitude. Until like the late '00s "telemetry" was, full stop, spyware (still is, for those of us who didn't shift our attitudes with the prevailing winds). I wouldn't say that responsiveness to user needs and desires has improved since then, in software design.
But what is the problem? That I can know that you press the print button? That you chose the Edit menu? I really don't see the problem. Please, explain, I really want to understand.
You don't see the problem of someone recording the actions you take using your own computer in your own home or office? It's like having a stranger sitting over your shoulder watching you. It's creepy and weird, and it's gross that people try to do it at all.
It's one thing to argue over whether basic user facing software like an image compressor or a text editor should have telemetry, but a web browser is one of the least controversial scenarios for telemetry I can imagine. It is constantly sending and receiving data on your behalf with hundreds or thousands of servers spread across the internet as a user agent. Your usage patterns - i.e. is it crashing, is the feature you're trying to use failing to work for some reason, is it rendering at a good framerate, is it running out of memory, are you having trouble finding the information you're looking for - are going to be incredibly complex and specific to you.
Significant bugs can affect only 1% or 0.1% of a browser's userbase but at Chrome scale or even Firefox scale that's like a million people. If you don't have telemetry it is REALLY hard to hear from those people about their problems and understand them. There simply are not alternative solutions that work half as well as opt-in (or opt-out) telemetry. People who say web browsers don't need telemetry are simply ignorant of what it's like to ship one and try to keep it working in the face of a constantly shifting environment - broken drivers, broken VPNs, malicious websites, malicious extensions, broken hardware, and users who are confused or tired or simply just bad at using software. No one is speaking on their behalf, you have to dig their suffering out of the data by looking at crash reports and performance metrics.
Shipping a web browser used by a million (or a billion) users means that you have a responsibility to do a good job. If your browser is not well engineered and reliable and responsive to users' needs that can result in data breaches or third-party server outages when your browser misbehaves or incorrectly channels user intent.
I'm personally a fan of making usage telemetry opt-in instead of opt-out, but browsers are a case where I don't opt out because I know how important the data is for browser vendors to make informed decisions.
This is of course different from sending your browsing history to Google, Microsoft, or any other company. I encourage people not to opt in to that stuff and not to sync their history/bookmarks/etc to those companies.
> It's one thing to argue over whether basic user facing software like an image compressor or a text editor should have telemetry, but a web browser is one of the least controversial scenarios for telemetry I can imagine. It is constantly sending and receiving data on your behalf with hundreds or thousands of servers spread across the internet as a user agent.
It's probably no accident that spying on users got popular just as this became the case. Constant network traffic while web browsing didn't start to become the norm until late in the '00s, either. If you weren't clicking links, you could often open Wireshark or sniff with Netcat and see nothing. Not from your browser, not from anything. Certainly ~nobody was collecting heatmaps of where you move your mouse, or firing a network request if you selected text. Or recording entire user sessions for playback, or so you can watch them live (god, those tools are creepy as hell)
The prevalence of "every app you use is a web browser now" is absolutely a catastrophe for user privacy and software reliability for this reason, IMO. Every tiny component now has a thousand moving parts that can spy on you.
> But what is the problem? That I can know that you press the print button?
When the internet was young, and most people were using dial up connections, just collecting the dates and times that a person was online and using a program was (and still is) a massive violation of privacy. Software "phoning home", even just to check for updates (collecting IP addresses, timestamps, and version numbers) was enough to get your software branded as spyware.
No software company needs to know which hours I'm awake, when I'm using my computer, which hours I work, which hours I use their program, how long I use their program, how long it's been since I last used their program, etc. It's intrusive, entirely none of their business, and it's insane that they all feel entitled to that kind of information.
If I print something, don't print something, or what the things I print are is also none of their business. Neither is what I'm printing it for, where I put the printout after I take it from the printer tray, or if I use tape or a thumb tack to secure it in place, but you can bet that if software could easily collect that data it would and somehow it would be considered impossible to write good software without that information.
From a privacy standpoint telemetry is always invasive, which is why I disable it any way that I can. Even without the privacy aspect telemetry is a bad idea. I don't want program updates that remove features just because I (and others) don't use them very often. I don't want updates that constantly shuffle the UI around according to how they think "most" people have been using it this week. I don't want my workflow disrupted every few months because it's uncommon. I don't want the way I choose to use the software on my device to influence how other people are expected to use it either.
Telemetry is much better when it's limited to reporting errors and bugs, but even that should be opt-in only.
2) Conduct user studies
How are companies that aren't software vendors and aren't able to spy on their customers able to do it? Did software companies not have good ways to do this before spying on their users?