Adrian Cochrane
So there's another app-analytics controversy in opensource, this time regarding Go Lang? (Yes I shared some pro-advocacy blogposts this morning)
I do agree: Ideally we won't have any analytics in free software.
But these controversies will keep springing up because this is how proprietary devs are used to making decisions, so whenever the next one moves over to opensource we'll go through this rigamarole yet again.
So I think its important to ask: What compromise is reasonable? Go gets close.
R. L. Dane :debian: :openbsd:
Not familiar with this particular brouhaha, but I think opt-in analytics with example reports are the way to go.
I thought the outcry over Audacity was hilarious*, given that even Debian has an (optional) package called popularity-contest that reports on what packages people have installed.
*The part where they were going to use GOOG (iirc) for the analytics was not funny, though. Poor choice. But it was still voluntary.
Adrian Cochrane
@RL_Dane I'm not a Go programmer personally, but here's the blogposts:
http://research.swtch.com/telemetry-intro
http://research.swtch.com/telemetry-design
http://research.swtch.com/telemetry-uses
Diane 🕵
I think users may appreciate some of the UI improvements that analytics can allow developers to figure out.
My feeling is an analytics system should allow the user to easily turn analytics on and off, provide a human readable summary of what is being collected, and provide access to the raw data that is being sent, and clearly specify how and what the analytics will be used for.
Ideally there would also be an open and transparent process for discussing what kind of information should be collected, how best to do summarization on local hardware, and at least try to do anonymization.
I think people are would be happier with analytics that are used solely for improvements, and isn't getting resold to anyone else.
Adrian Cochrane
My solution: Have our own differential privacy app-analytics service! Clients would inject uniformly-sampled random noise into this analytics so that its only useful for aggregate statistical analysis, and I'd upload via UDP so it works well across any UI toolkit. By default gather no data.
Terms-of-service would require opt-in, & the gathered histograms would be made public with outliers removed.
Maybe make this a freemium service so proprietary software is a little less disgusting?
R. L. Dane :debian: :openbsd:
Couldn't follow all of it, but it was fun to skim.
I think they make some good points. The need for telemetry in a *language* is bizarre, but an understandable consequence of the shift away from stable software and languages to a constantly shifting platform.
I for one, kinda miss stability. I don't mean that things are always crashing, but rather that things are changing so quickly.
...
R. L. Dane :debian: :openbsd:
...
Example: I tried to install and use Virtscreen to turn a surplus old iPad into a second laptop monitor, but found that it was unusable. The last commits were five years ago, from what I can tell. Five years, and whatever python packages/libraries it relies on have changed to the point that it renders the project useless without further refinement.
That... is just nuts.
Adrian Cochrane
@KryptykPhysh https://floss.social/@alcinnz/109831337176853802
KryptykPhysh
@alcinnz This one has slipped me by. Pointers (ha!) to the controversy, please?
LisPi
@RL_Dane @alcinnz One of the many reasons I've cast my hat into the #CommonLisp ring.
Martin Atkins
@RL_Dane @alcinnz as long as old versions of all involved software remain available, you can keep your stability by running those. And in open source it's far more likely that older versions will be available.
In my experience the big trouble starts when people want to mix and match new and old. Everything is either explicitly or implicitly designed for the environment that exists at the time it was created, and we cannot predict the future.
Martin Atkins
@RL_Dane @alcinnz for example, people want to run software on macOS running on Apple Silicon, and that platform is very young.
Most software in the world was not designed with that platform in mind. If you want to use Apple Silicon then you probably need to use new software.
If you want to run old software then you ideally need an old OS running on an old CPU architecture. But in practice that's not what most people seem to want.
Martin Atkins
@RL_Dane @alcinnz meanwhile the retro computing hobbiest community is making heroic efforts to do things that most computer users would find pointless, like rebuilding an Amiga in an FPGA. I love that this stuff exists, but it does also demonstrate that keeping old stuff running is _real work_, and somebody has to care enough to do it.
グレェ「grey」
@alcinnz Yeah, I dunno. Personally I vehemently oppose the sort of telemetry by default approach that Homebrew uses as mentioned in TFA.
MacPorts in contrast (which is an older project, so old it began as DarwinPorts co-founded by jkh who also co-founded FreeBSD) does have mpstats for those who want to opt-in to telemetry, but spyware should NEVER be a default, it should ALWAYS be a choice.
There are other *saner* methods too e.g. running AWStats or Webalizer, not Google Analytics.
@RL_Dane
グレェ「grey」
@apparentlymart Apple Silicon may be "young" but ARM is from the 1980s.
C, is from the 1970s and UNIX was ported from PDP11/20 assembly to C, to facilitate portability.
Which seems to have worked remarkably well now, for decades.
Problematically: in the 1970s: most people using computers also programmed. In the 21st century most people using computers, just use them. I can't fathom why anyone would use a computer who doesn't program, yet here we are, and it's horrible. @RL_Dane @alcinnz
グレェ「grey」
@apparentlymart Keeping old stuff running does take a lot of effort (I write as someone who was previously employed by the Museum of Art and Digital Entertainment a 501c3 nonprofit wherein I was often asked to recreate prior art to help attorneys for Fortune 25 invalidate spurious patents).
Unfortunately, it doesn't tend to pay well, if at all. Because: monetization in our society hyper fixates on the latest and greatest unless you happen to own a Stradivarius, or Amiga. ;) @RL_Dane @alcinnz
R. L. Dane :debian: :openbsd:
Yeah. In this case, it would probably work fine on something like Debian where versions are held for a couple years at a time.
But on a rolling release distro, things not maintained tend to break.
This is absolutely not ideal.
I'd deal with balooning exeutables to get everything statically linked, but that then comes with security problems.
R. L. Dane :debian: :openbsd:
Very true. It's amazing that that cottage industry is thriving the way it is.
I love it.
ozamidas
@alienghic @alcinnz while i do not necessarily disagree with you, it was through the use of analytics that Mozilla took away the RSS features from Firefox.
Honestly, the European Union got most of this right, maybe go devs should follow that.
R. L. Dane :debian: :openbsd:
@byterhymer @apparentlymart @alcinnz
Not only that, but the iPhone has dumbed people down to the point that they just accept any blasted thing.
They'd give away a kidney if it was in the EULA of a cat video app. 🙄
Martin Atkins
@byterhymer @RL_Dane @alcinnz indeed, the architecture is not young but the combination of macOS and ARM is, and it is that pairing I was referring to.
Open source software written for MacOS on x86_64 is typically relatively easy to port as long as it's written portably, but proprietary software is not so easy.
My general point though: making old software run on new platforms isn't necessarily free. And keeping old platforms working isn't free either.
０７６萌ＳＮＳ is a social network, courtesy of ０７６. It runs on GNU social, version 2.0.2-beta0, available under the GNU Affero General Public License.
All ０７６萌ＳＮＳ content and data are available under the Creative Commons Attribution 3.0 license.