Unpopular opinion but, is opt-outtable telemetry that bad?
Of course using tracked behaviors for add targeting is bad (which is why my opinion is irrelevant to the article) but using telemetry to do AB-tests, UX-tests, finding bugs, etc... what’s really bad about it?
Every comment about telemetry in HN reads something like ‘Telemetry with opt-out is bad, so GitLab is bad, VSCode is bad, brew is bad, etc...’ but personally I really can’t see the problem.
For people who can’t believe the company enough for processing one’s data can turn off telemetry right? The fact that one is using the application from one company (for example, VSCode) implies that they trust the company (in this case, MS). If one doesn’t, one can manually turn it off.
I would argue that opt-in telemetry is the worst you should think about.
To improve your software, there are other means than mass surveillance of your users (e.g. lab-tests). And if you really need a lot of data, ask who wants to participate and also offer a solution that lets them participate for a limited amount of time.
In my experience, you can find a lot of problems with lab tests and when your team disagrees about some findings, quantitative tests can help you find the correct answer. But tracking all your users constantly isn't necessary to reach a good user experience.
Yes it is bad. Most of it relies on tools provided by advertising & stalking companies like Google Analytics, so they have an incentive to also use the data for themselves, and correlate it with other data they collect about you when you use their services directly. This also means these companies (even if not involved in advertising directly) have an insane potential in tracking users across all apps that happen to use their services, making them a very juicy target, including for state actors.
> AB-tests, UX-tests
I don't want to be a lab rat. If you are testing things then please make it clear and let me see both options and I will happily give feedback.
> finding bugs
What if the software crashes while processing confidential data and part of that data is contained in the core dump that you just sent off because you don't respect your users enough to ask them?
> Unpopular opinion but, is opt-outtable telemetry that bad?
In its current incarnation, yes.
From what I've seen, telemetry has simply replaced quality assurance and proper design, and I don't like it.
It's hard to put it into words, exactly, but I'm old enough now to realise that using primarily statistics to drive high-level decisions inevitably leads to a general crapification of everything. It naturally leads to dark patterns, to gamification, to making things addictive... sorry... "engaging". It sneaks in prejudices, and ossifies the status quo. Worst of all, it tends to drag everything down to the lowest common denominator, to the left side of the bell curve. Mass appeal is appeal to the average and below-average person.
This type of viewer-feedback design is why every Hollywood movie is now so formulaic. Must have orange-blue contrast poster. Must have a love story. Must have uplifting moment. Must have plucky ethnic characters. It has become revolting slop for the masses to consume in bulk.
Software design is no different. I want quality, I want inspired software that is cohesive, fit for purpose, and will teach me something. Instead it is being purposefully lowered to the level of the average, computer-illiterate person in order to be more "accessible".
As a random example, I was infuriated when Microsoft announced that they're deprecating the Windows 7 version of the backup tool from a future Windows 10 release because their telemetry shows that most home users don't use it much.
Who. Gives. A. Fuck. what some grandma does or doesn't do with their worthless PC and its virus-riddled hard drive with no valuable data on it?
Do you? I don't. Microsoft shouldn't.
But oh no, their telemetry says Grandpa and Grandma don't use VHDX-based image backup enough, so it's go to go.
I use it regularly because it can back create disk-image backups of a running machine at a button press. It reads the source disk linearly block-by-block, and hence I can use it to back up -- or restore -- the 450 GB of tiny source code files on my work laptop in just 15 minutes! If my laptop dies, I can just attach the backup disk to another computer with Hyper-V and directly boot the latest image file. No need to even restore.
This is amazing engineering! It reuses the built-in Volume Shadow Service (VSS) engine to take the backup, so it's practically free code for Microsoft, but... nope. Sorry. The "people" have spoken. Less than 0.1% of the user base clicked this icon in the last 12 months, so clearly I'm wrong and the masses are right.
I would be okay with it if it was properly anonymized. I'm talking for instance some peer-to-peer mixing scheme, not just a privacy clause by which they swear they won't use the data to identify the particular users. There are technical solutions to privacy-conscious telemetry, but of course the non-trivial cost of implementation, and perhaps more importantly the loss of valuable data for the company means it's almost never implemented.
> I would be okay with it if it was properly anonymized.
MS might properly anonymize it, CandyCrash as well, then antivirus company also does it properly, but when it's all combined nothing says it won't be properly de-anonymized, specially if the data is sensitive and valuable.
If the telemetry information is only internal to the application (generic UI interactions, no specific IDs or tags), I don't see how combining sources could de-anonymize it.
Yes, it is that bad.
Dark pattern opt-in tomfoolery is bad enough.
The current 'good luck finding your way out of this maze of never ending twists and turns' opt-out horror scenarios is 100x worse, and that is after you found out about all the tracking and reporting shenanigans and the existence of the opt-out in the first place.
I use Windows 10 pro out-of-the-box on all my machines. I change few default settings. Seeing a tile for the latest Microsoft game in the Start menu just doesn't bother me. I like Cortana. I just checked on the weather.
I note that Macs have an unremovable Siri, but nobody here is upset about that.
One of the things that I didn’t see anyone else mention is that opt-in telemetry tends to annoy a certain group of people, who will turn it off and then you’re not getting anything from them. Then you make decisions that are non-representative.
I, personally, love telemetry and always enable it everywhere. I want developers to know about my habits and consider it when deciding, for example, which features to delete.
I, personally, enable telemetry everywhere it's disabled by default, or even just when I'm asked about it. I turn it off if it's on by default and they don't say exactly what they're collecting.
Unpopular opinion but, is opt-outtable telemetry that bad?
Of course using tracked behaviors for add targeting is bad (which is why my opinion is irrelevant to the article) but using telemetry to do AB-tests, UX-tests, finding bugs, etc... what’s really bad about it?
Every comment about telemetry in HN reads something like ‘Telemetry with opt-out is bad, so GitLab is bad, VSCode is bad, brew is bad, etc...’ but personally I really can’t see the problem.
For people who can’t believe the company enough for processing one’s data can turn off telemetry right? The fact that one is using the application from one company (for example, VSCode) implies that they trust the company (in this case, MS). If one doesn’t, one can manually turn it off.