Ask HN: Should open-source projects allow disabling telemetry?

27 points by igorzij 2 years ago | 29 comments

We just had a user submit an issue and a PR to revert the changes we made earlier that remove the option to disable telemetry. We feel like it’s a fair ask to share usage data with authors of an open-source tool that’s early in the making; but the user’s viewpoint is also perfectly understandable. Are we in the wrong here?

https://github.com/diggerhq/digger/issues/1179

Surely we aren’t the first open-source company to face this dilemma. We don’t want to alienate the community; but losing visibility of usage doesn’t sound great either. Give people the “more privacy” button and most are going to press it. Is there a happy medium?

wkat4242 2 years ago |

Telemetry should always be opt-in. Especially for FOSS.

We shouldn't have to pay with our privacy.

Be aware that the EU is working to make opt in mandatory too. But you won't be their target of course, it'll be the truly evil companies like Microsoft, Meta and Google.

> Give people the “more privacy” button and most are going to press it. Is there a happy medium?

This should really tell you enough. If you already know users don't want to share the info if you would ask them, you're doing the wrong thing by withholding the option.

The big techs spend so much money developing dark patterns exactly for this reason.

viraptor 2 years ago |

"We feel like it’s a fair ask" - exactly - it's a fair ask, so the user should decide. Nobody can force you to change your mind, but if you don't, prepare for forks with the option disabled.

Also, the issue is not really about that - it's about the submitted data not being anonymised. That's the bare minimum you should be doing regardless of the information being opt-in/out.

asadotzler 2 years ago | |

The user can decide, by not using the software.

viraptor 2 years ago | | |

Sure, but for opensource it also means the user can maintain a forked version. It's up to the project to decide if they prefer to fight over privacy, have a fractured community, and possibly custom patched packages in some distributions - or do they explicitly ask about telemetry instead.

And in practice, do you gain enough from active collection to justify the fight? What about instead searching GitHub for digger being used in the public pipelines and building stats from that?

nusl 2 years ago | | |

The software should give you an immediate notice before you use it at all, then, and make you agree explicitly. That way they can say no. Otherwise you’re just abusive.

calamari4065 2 years ago | | |

That's not a choice, that's an ultimatum.

andymoe 2 years ago |

You are in the wrong PR wise probably. Generally I think providing a way to opt out has become the path many projects take.

Anyway, their main beef appears that you are not properly anonymizing the telemetry.

igorzij 2 years ago | |

Thanks! This is helpful. None of us at Digger have led or maintained significant OSS projects before - so discovering things as we go

bcon 2 years ago |

I've [argued](https://benconrad.net/posts/230717_fundingOpenSource/) that oss developers need to have real world use information in order to ship good software. Contra some here, I think oss developers that don't understand how users use their software can't maintain or improve it.

justinclift 2 years ago | |

Bear in mind that some of the people reading this comment are OSS developers, potentially even ones who write software you're personally using. ;)

> I think oss developers that don't understand how users use their software can't maintain or improve it.

For trivial improvements, that's definitely not the case.

For example, if someone files a GitHub issue saying "please add new menu shortcut XYZ". That's the kind of thing which doesn't directly need in depth understanding of how users are using the software.

More major features though will indeed benefit from understanding how users use it. That's not really a case for telemetry though, rather it's a case of having a wide enough user base that users ask for changes/features/fixes themselves.

Sometimes the users asking for changes/features/fixes is us ourselves. ;)

nusl 2 years ago |

I would straight-up uninstall your project and take any measures possible to remove myself from your platform (if any) if I discover an underhanded change like this.

A medium? Ask people. FOSS is different. There is a level of expected transparency and trust along with it that you can’t get back when broken. Changes like this are a very easy way to break that trust. Trust is everything here.

I expect to be tracked by commercial tools for the most part, but I still try to control how.

How would you feel if a trusted tool changed something you deem important without telling you? Generally a response is feeling betrayed.

The fact that someone skilled enough took time out of their lives to make this PR means that it matters enough to seriously reconsider.

eternityforest 2 years ago |

I think they should always allow it, just simply because... if you don't you might lose users, they might fork it and make a privacy-first version, then we get more fragmentation.

FOSS has so much fragmentation trouble already.

mtmail 2 years ago |

You should allow disabling it. The wording "allows collecting anonymised usage and debugging data" in the documentation doesn't make it clear enough that the data is send to the company servers.

nonrandomstring 2 years ago |

You are very definitely "in the wrong".

If you want to discuss this in more depth I'll be happy to elaborate and give you ethical guidance, but presently that's probably just adding to the noise.

igorzij 2 years ago | |

Thanks! I'm glad we asked. Felt uneasy either way. Also would appreciate any pointers, perhaps someone you know has written on ethics of it?

nonrandomstring 2 years ago | | |

YVW. I'll have a think about some other sources, especially plain speaking non-academic takes on the ethics that help developers see the issues. For now this one is a good general overview [0].

The big one with telemetry, is unintended side effects due to correlation and deanonymisation - which is actually dead hard to anticipate - very easy to get wrong like rolling your own cryptography :)

The other, around consent and defaults, is that even if your telemetry is perfectly anonymous, benign and beneficial to the end user, you may trigger a security alert and over-zealous investigation and reporting. This can have a massive impact on your reputation, as happened to Audacity. It's really not worth taking the risk.

Hope that helps.

[0] https://www.emerald.com/insight/content/doi/10.1108/S2398-60...

vunderba 2 years ago |

You should allow for the user to disable it preferably during install. Just try to allay their fears by communicating that it is anonymized and how it's important to help improve the software by providing this feedback.

Another option that I've seen is to inform the user that there is telemetry, but make the option to disable it some kind of configuration by editing a JSON file. This way you still allow for the option to disable telemetry, but add a small amount of friction to do so.

_mlbt 2 years ago |

I recommend giving your users the option to allow telemetry or not (as others have said, preferably by opting in). Explain prominently within your software very clearly what telemetry you are collecting and for what purpose.

Since this software in question is open source, if you don't allow the disabling of telemetry I guarantee a fork will appear in the near future where it has been disabled altogether.

Ask yourself if this is a hill you wish for your project to die on?

curious_curios 2 years ago |

The go project team did a deep dive into this last year. There was a pretty interesting blog post[1] and discussion[2].

1: https://research.swtch.com/telemetry-intro

2: https://github.com/golang/go/discussions/58409

RyeCombinator 2 years ago |

It should be opt out by default and well communicated.

wkat4242 2 years ago | |

Opt out by default? You mean opt in.

Just ask the user on the first launch.

Rapzid 2 years ago |

What does "wrong" mean in this context?

It's your project you can manage it however you want. I'd recommend communicating this clearly at least but otherwise listen to your users(in aggregate) and make your decision.

KomoD 2 years ago |

> Should open-source projects allow disabling telemetry?

All software should, regardless of it being open-source or not. It should also be opt-in.

Not even anonymizing the data is also a MASSIVE red flag

calamari4065 2 years ago |

> Are we in the wrong here?

Unquestionably.

What I do with your software is quite frankly none of your business. You do not have the right to know what I do with my hardware. Your metrics are not my concern.

FOSS is a very personal thing. Try framing it as a conversation between you and your users:

"Hey, we work hard on this and would really appreciate it if you share telemetry so we can focus our efforts better"

"This application now collects telemetry. You cannot turn it off. You want privacy? Too bad."

The difference is asking your users to share data and demanding they give up their privacy on your whim.

> Give people the “more privacy” button and most are going to press it.

You should have a good long think about what this means and why removing privacy options is making your users upset.

When I implement telemetry in a FOSS project, I go well out of my way to do it in the most respectful way I can. This usually takes the form of a detailed list of what is and is not being collected, along with why I need it. During first run, there is either a very prominent notification or a full on blocking step where the user must choose whether to enable telemetry. This includes text explaining why I want telemetry and a link to the documents.

My philosophy is that users are not idiots and they are not children that need coddling. I make my request, give them the information, and let them make their choice. That's it, end of transaction. The user made a choice and the matter is totally out of my control.

Anything less is not respecting your users.

asadotzler 2 years ago |

If you allow opting out throw away the telemetry. At your scale your results will be garbage with opt out. Don't even bother.

eimrine 2 years ago | |

I don't want to use your software.

kiicia 2 years ago |

Why would you require any telemetry? Just let users use your Foss app as they see fit...