Telemetry and crash reports can be done right so you should not hate the concept but the bad implementation.
First step is that it should all be optional and it should be disabled by default or have it as an option you are forced to decider if yes or no. You should also be informed when data is sent and what is sent and have the option to deny. Like if app X crashes I send the crash report but if app Y crashes I won't send it.
Also about telemetry if all power users turn it off then you can have situations where developers will drop a feature because they don't use it and there is no data to show that more then 12 people are using it.
> Also about telemetry if all power users turn it off then you can have situations where developers will drop a feature because they don't use it and there is no data to show that more then 12 people are using it.
Yet another reason not to put telemetry in your software.
Why? So some developers and jerks can claim that you are a snowflake if you used the feature they removed? If there was data at least you can resign that you are really special and you were in a select club but without the data the vocal developer with big ego will claim whatever he wants.
When something important gets removed (which is happening at an alarmingly increasing pace across the board), it doesn't matter at all to me why the feature was removed. What matters to me is that the software has become less suitable for my use.
>Yet another reason not to put telemetry in your software.
Can you explain a bit more your way of though? How should you decide what to support if not having the information from your user?
The only other alternative is to ask, and then only the one that answer will get supported, which sure is great for the one that answer, but that may means a much bigger audience that you won't support even though they need to be supported.
> How should you decide what to support if not having the information from your user?
There are many well-known and largely perfected methods of getting that information without the use of telemetry. The thing that general telemetry really brings is cost savings and convenience. But it also brings a peculiar kind of blindness as well, since companies largely stopped doing product research that isn't telemetry.
>There are many well-known and largely perfected methods of getting that information without the use of telemetry.
Can you link to a few of this many known methods? I am interested in ones that would fit open source projects that have no budgets for research. Thanks
Just look at how product development worked before telemetry was a thing that was possible, and how its done in other industries. This isn't hidden information.
Telemetry is 100% not a viable fit for an open source project. A very significant proportion of your potential users will either keep it disabled or if that is not possible (which would be illegal in the EU anyway) not use your software.
> A very significant proportion of your potential users will either keep it disabled
Then they made the conscious choice that their usage won't get known and thus be aware that another way of communication will have to be used.
> Just ask your users.
"A very significant proportion of your potential users" won't tell what they want, or won't even know what they want. The one that will be the most vocal won't necessarily be the voice of reasons either.
... nothing wrong with what followed? Why did you just quote the beginning when I cite a pretty good argument afterward and you don't even argue with it? We are on Hacker News please, be more thorough and defends your point.
So? Just because we happen to be on Y Combinator's corporate propaganda machine doesn't mean I'm any more compelled to respond to things about which I care not.
You're saying that as if the point of the software is to support the software, or report good numbers. It's not. The user should be the focus.
Opt-out and hard-coded systems immediately put you no the back foot. You make people question why? and is it worth it? If all you've got it reporting numbers of users, you can go do one. The risk of genuine private data leaking out is not worth your marketing numbers to me.
The Amazon link was much more egregious though. You were sending all Dash searches through to Amazon. This was unacceptable and poorly explained. I'm sure some people liked it, and they honestly could have made the integration a better trade for the user (play videos off Prime Video, etc), but only as an opt-in feature where you can see the ramifications.
> The Amazon link was much more egregious though. You were sending all Dash searches through to Amazon. This was unacceptable and poorly explained. I'm sure some people liked it, and they honestly could have made the integration a better trade for the user (play videos off Prime Video, etc), but only as an opt-in feature where you can see the ramifications.
You do realize that no version of Ubuntu was ever released with that feature? Sure, the Amazon scope was available for a while but you had to opt in in all shipped versions. Even if you had, Canonical turned their servers off in 2017 so it wouldn't have done much except time out.
It's easy to get upset about stuff people have made up in echo chamber "news" sites.
The feature article is about the link to open Amazon in the browser, not the Amazon scope in the Dash.
> You do realize that no version of Ubuntu was ever released with that feature?
Not sure that is true - I seem to remember that it was enabled by default in a couple of releases (I think somewhere around 12.04?).
> you had to opt in
> It's easy to get upset about stuff people have made up
I raised this issue early in the Lens development phase. The response from the mailing list was (paraphrasing) "well, people can always uninstall the package if they don't want the feature" (at which point I then opted out entirely).
People need to understand that large scale software not having telemetry borders on malpractice if it interacts with the internet. If your software is crashing on a large scale - or worse, causing problems for other people's servers or taking out users' computers - you need to know, and you need to fix it fast. You can't just wait days or weeks to get enough complaints about it from people savvy enough to file bug reports.
Look no further than how a recent Chrome update almost completely broke the software for Citrix users. Their telemetry (somehow) was not tracking this so it took days to get a fix because IT/ops people had to figure out that a Chrome update caused the bug and then escalate through chromium bug reports. Now imagine that happening any time a crash or major bug makes it into connected software...
For stuff that just runs offline, like Excel or whatever, it's super justified to not want telemetry in it if that's the tradeoff users want.
Incidentally this is basically the reason why all the major browsers are on rapid forced (unless you're in the enterprise) updates - the risk of having users stay on older versions with security holes is just too high because it ends up leading to massive botnets. I hate it, but hundreds of millions of people use those apps.
What's wrong with the little popup that asks you if you want to report a problem every time a crash happens? Is there really any problem with letting the user choose on a case-by-case basis?
If it's happening so frequently that the popup becomes annoying, you've got bigger problems.
Providing the choice is always OK and often appropriate. The problem is when people argue that software shouldn't have telemetry or automated updates at all, because then what you get is software that respects the user's desire at a surface level - no telemetry, no automated updates - while failing to respect deeper aspects of their intent, like "I don't want to get added to a botnet", "I don't want this software to crash", and "I don't want features I like removed in future updates".
You can always decide to go elsewhere of course, and software developers have the right to decide personally not to put that stuff into their apps - none of my personal projects have telemetry and most of them don't have updaters.
The problem with web browsers is that you really can't. There are really only 2 options now, and the web has become so ridiculously complicated that creating a reasonable alternative browser is nearly impossible.
Because people rarely click Yes, even if it's rare.
Plus it means little to the end user, and outside of HN, Slashdot, or other tech forums, most folks don't even know what telemetry means in an OS context. They click No, and move on.
> Because people rarely click Yes, even if it's rare.
In my observation the precise opposite is true; your average user will immediately click "Yes" without reading in the hopes of making the error go away.
So the solution is to deprive them of their privacy for your own benefit? I don't blame them for clicking "no" if they don't know what it will be used for.
That's fine, if it is opt-in and clearly spelled out. Some of us do confidential work and leaking URLs visited and field contents to parties not subject to NDA is absolutely not acceptable in any context.
Yeah, a coherent privacy model, disclosure etc are all super important. It's just irritating to see people act like software of this type can rationally be shipped without any telemetry to a big group of users who aren't compiling it from source.
Naturally a ton of work has to go into properly designing telemetry for something like Firefox or Chrome - not only is disclosing URLs a no-go, but you need to make sure your telemetry isn't unintentionally exposing details that are identifying in some other fashion, like on-stack data from a crash report or that sort of thing.
This is one of the strongest arguments against an Internet monoculture. Shame about all those alternative browsers dying to the point that now if you're working on one, CloudFlare/Google/etc are likely to just block your browser entirely.
Everyone's using the same massive-scale app and now that it's a central point of failure it's wildly irresponsible to not load it down with telemetry. It sucks. If it were instead a healthy market of like 4-8 different browsers people were using, a major issue in one of them wouldn't necessarily be a crisis because people could just boot up a different browser and the bug would only impact say 5-25% of a website's visitors at most instead of The Majority Of Them.
I agree and this is why I said good implementation can exists. I never implemented telemtry but I implemented once crash reporting. Before I added this feature the users had to contact support and tall us what is wrong, then we would send email back and forward asking about OS , application version. asking for screenshots, asking to enable debug mode and try again and then screenshot the error and send it to us. Because i was the one doing the customer support too I noticed this is very repetitive and very demanding for users (even explaing a user to find a log file in %appdata% in windows is a complex process) so I simplified it as much as possible, a crash or the user could open the report a bug popup, you would see exactly the data that is sent to me, it was plain text info like OS and app versions and a stack trace and other logging data and you had the option not to send it to me but if you sent it then I have the stack trace I have the info and I could put an update to fix teh bug in a day rather then we spend time emailing each other questions.
> or have it as an option you are forced to decider if yes or no.
Bonus points for an implementation that comes with a few unbiased examples about why one might want to disable and why not.
You are working under draconian legal secrecy requirements or on an exceptionally narrow/expensive uplink? Disable. You worry about secrecy but not enough to go 2FA, you occasionally wish for a bigger pipe to the internet? Disable if you insist, but it would be nice of you to contribute data.
If you are working with such parameters, one would hope you wouldn't have to be explained that telemetry should be disabled. Not all apps will be so kind as to give you examples...
The purpose of the examples would not be to keep those out of tracking whom you'd expect to deactivate (because you are right, it becomes entirely their responsibility once they are actively presented the option), it would be to encourage trust with those who don't really need to be that level of paranoid (but who often are). Much better than the usual "trust us, everything is so incredibly safe that we don't even know why we have the option to deactivate". Most users will take the "incredibly" at face value and not believe it, and they would be right because nothing is ever that safe. But a risk model like "theoretically there could be a bug in that totally innocent feature use counter that accidentally included a random heap sample in the upload that could include sensitive data" would be acceptable to many. Could a liar smooth-talk you with those examples into pressing accept? Sure. But that bad actor could also ignore the setting or not show an option at all.
>Also about telemetry if all power users turn it off then you can have situations where developers will drop a feature because they don't use it and there is no data to show that more then 12 people are using it.
Firefox and the RSS situation. Mozilla did not even think 1 second that this might occur.
Live Bookmarks were no substitute for a proper RSS reader. I've been using both RSS and Firefox continually since at least 2005, but I quickly stopped using Live Bookmarks in favor of Google Reader (RIP).
Telemetry and crash reports can be done right so you should not hate the concept but the bad implementation.
First step is that it should all be optional and it should be disabled by default or have it as an option you are forced to decider if yes or no. You should also be informed when data is sent and what is sent and have the option to deny. Like if app X crashes I send the crash report but if app Y crashes I won't send it.
Also about telemetry if all power users turn it off then you can have situations where developers will drop a feature because they don't use it and there is no data to show that more then 12 people are using it.