pkgstats: contribution was never that easy

Chris.Giles · 2009-02-16 09:47:28

Allan wrote:

I think that you need to learn to use the edit button....

Thanks for the tip. Maybe they forgot to cover this during my computer systems engineering degree. It's possible that they only do so at the master's level.

Allan wrote:

And given you repeated yourself:
http://bugs.archlinux.org/task/12991#comment39515
pkgstats is an Arch specific construct and pacman aims to be distro agnostic. So pacman will not be calling pkgstats.

I thought you'd figure out that I posted duplicates here, just in case nobody was still monitoring the feature request ticket. I'm guessing you instead wrote me off as being an idiot, plain and simple.
You've made a good point about the need to keep 'pacman' distribution-agnostic. I reckon others would see the benefit in having the 'pkgstats' framework present in 'pacman', even if it's disabled by default in the source archive. There's nothing stopping 'pkgstats' from also being unspecific to Arch Linux.
If 'pkgstats' is rewritten in C and compiled into 'pacman', it could more safely use the computer's MAC address to determine uniqueness. This would provide more accuracy than using the IP address currently does.

crouse wrote:

Not to mention, some users would prefer to keep (for whatever reasons) everything private. That should be their choice as well.
Sending data back, without the consent of the user, nope, not in favor of that myself.

Yes, I thought about this afterwards. Perhaps 'pacman' could ask the user for permission to submit the information about the packages, every hundredth time it's executed. If the user rejects the request, 'pacman' could just instruct 'pkgstats' to submit blank information and the database would at least be aware that another Arch Linux user exists.

Allan · 2009-02-16 09:55:59

Chris.Giles wrote:

I'm guessing you instead wrote me off as being an idiot, plain and simple.

Don't worry... It isn't just you! Seriously, I was just being blunt because this was brought up many times when pkgstats was first released. But I see you are new around here so you probably had not seen this.

Anyway, welcome to the forums. We are normally a friendly bunch

phrakture · 2009-02-16 17:42:34

Chris.Giles wrote:

If 'pkgstats' is rewritten in C and compiled into 'pacman', it could more safely use the computer's MAC address to determine uniqueness. This would provide more accuracy than using the IP address currently does.

I'm fairly certain that pkgstats will never get merged into pacman. pacman is a utility to manage packages installed on your system, and not a utility to inform the Arch devs about what people have installed. Two utilities make MUCH more sense then one giant integrated blob that does 40 disparate things. See http://www.faqs.org/docs/artu/ch01s06.html

dyscoria · 2009-02-16 21:37:51

Chris.Giles wrote:

crouse wrote:
Not to mention, some users would prefer to keep (for whatever reasons) everything private. That should be their choice as well.
Sending data back, without the consent of the user, nope, not in favor of that myself.
Yes, I thought about this afterwards. Perhaps 'pacman' could ask the user for permission to submit the information about the packages, every hundredth time it's executed. If the user rejects the request, 'pacman' could just instruct 'pkgstats' to submit blank information and the database would at least be aware that another Arch Linux user exists.

Ahem, I believe that would still be classed as sending data back

haxit · 2009-02-16 21:47:23

Yay ^^
I am blocked. I am so cool.

Chris.Giles · 2009-02-17 07:29:18

Thanks for the replies. You've all made good points, except maybe for 'haxit'.

You guys are the experts and I don't wanna be pushy, but I just think that enriching 'pacman' with an optionally disabled 'pkgstats' would further its reputation amongst the Linux community. It'd also be nice to get a larger sample of who's using what, as the majority of users aren't going to whine about boring old "privacy concerns".

We might also wanna think about pinching some of Conary's features, as seen in Foresight Linux. For example, our currently broken 'UseDelta' option could be superseded by Conary's incremental (RCS) updating mechanism, even at the package level.

haxit · 2009-02-17 20:15:34

Chris.Giles wrote:

Thanks for the replies. You've all made good points, except maybe for 'haxit'.

Sorry, my reply was for the original post

fukawi2 · 2009-02-17 22:12:24

Chris.Giles wrote:

You guys are the experts and I don't wanna be pushy, but I just think that enriching 'pacman' with an optionally disabled 'pkgstats' would further its reputation amongst the Linux community.

I have to agree that this isn't the "Arch Way". (Someone had to say it! ) I don't see how "phoning home" would raise the profile of ArchLinux in a good way.

As has been mentioned, pacman is a Package Management tool, not a Package Management and Developer Reporting tool. Linux's power lies in have many small tools that are each very good at their own little thing, not few monolithic tools that are jack of all trades and master of none.

EDIT: I could see a compromise where pacman has an optional dependancy on pkgstats to help raise awareness of the package among the users (those who use Arch without reading the forums, news etc or who just weren't around when pkgstats was first released)

My 2.2 cents (GST inclusive in Australia)

Last edited by fukawi2 (2009-02-17 22:14:10)

Chris.Giles · 2009-02-18 09:28:29

fukawi2 wrote:

I could see a compromise where pacman has an optional dependancy on pkgstats to help raise awareness of the package among the users (those who use Arch without reading the forums, news etc or who just weren't around when pkgstats was first released)

Yeah, I think having 'pkgstats' as an optional dependency for 'pacman' would be a good minimum. It would make more Arch Linux users aware of 'pkgstats', but there would still be many who'd ignore it and the statistical accuracy would therefore be lower.

If 'pkgstats' becomes a necessary dependency for 'pacman', it could still be optionally disabled in the configuration file or rejected by the user at every hundredth runtime. I don't think this approach would be un-Archie, especially if the 'pkgstats' functionality is disabled by default in the source archive for other distributions.

dyscoria · 2009-02-18 09:57:45

Chris.Giles wrote:

rejected by the user at every hundredth time

I must disagree. Sure pkgstats is useful, but to pester the user at all post install (let alone more than once) would be very un-Arch-like, and i'm fairly sure many distributions would share the same view (i know Fedora at least only asks once at install).

The only time a user should ever be asked to provide package statistics is on first install. After installation, providing statistics should be a completely voluntary activity taken by the user. Encouragement to do so on the website or forums is as far as a distribution should go. If an Arch user does not want to provide statistics then he doesn't want to be pestered _at all_ even if it's every thousand times that pacman is run.

Chris.Giles · 2009-02-19 10:06:07

I disagree with all of that, but I guess it returns us to the ubiquitous simplicity versus functionality debate.

Whatever happens regarding this issue, Arch Linux will remain my preferred distribution by a large margin. Well done everyone!

Radek · 2009-04-16 05:12:55

What do you think about adding information about pkgstats to Wiki, at the end of "Beginners Guide"? I think that would be good idea to place there something like "You have finally installed Arch! Don't forget to tell developers which applications are important to you!"

Xyne · 2009-04-28 18:59:39

Could you please add the link to the results page to the original post? It's difficult to find on google and most users won't dig through 4 pages to find it in this thread.
http://www.archlinux.de/?page=PackageStatistics

Svenstaro · 2009-05-02 22:21:40

+1 for adding the link to the original post

jowilkin · 2009-05-08 13:48:41

Chris.Giles wrote:

the majority of users aren't going to whine about boring old "privacy concerns"

Privacy concerns are not boring OR old. Reporting information about users is most certainly something that needs to be opted into, not opted out of. Especially not every hundred times I use Pacman, if I don't want to submit stats the first time it asks, it should not pester me about it.

Chris.Giles · 2009-05-08 22:34:06

IMHO, users who adopt attitudes like that help to prevent Arch Linux from becoming even more efficient.

If all identifying information is encrypted during transmission and on the server, users should not be afraid to report their packaging statistics.

SamC · 2009-05-08 23:01:02

There are some people, certainly, who do not want their information sent to what may be a rather unknown party. Privacy concerns are a very real thing, and encrypting the stats would not change the fact that the information, anonymized or not, is leaving your hands.

Perhaps pkgstats could be integrated into pacman, but the only way that would satisfy the concerns voiced here would likely be a default setting of off. The only way of turning it on would be adding a line or two to pacman.conf, similar to ILove*****. (Can't go spoiling anything, can we?)

fukawi2 · 2009-05-08 23:34:41

Chris.Giles wrote:

IMHO, users who adopt attitudes like that help to prevent Arch Linux from becoming even more efficient.

While that may be true (I'm not going to argue it one way or another), the facts remains that unique information pertaining to the user, anonymized or not, is being transmitted to an unknown/untrusted third party.

Another aspect is that forced data transmission like that could be a violation of company policies, which may inhibit the adoption of Arch in corporate environments.

linkmaster03 · 2009-08-16 17:49:01

I just submitted my data. Great idea guys!

mutlu_inek · 2009-08-16 20:10:56

AFAIK, pkgstats is discontinued for the time being.

Allan · 2009-08-16 20:23:05

Not really... I believe Pierre reset the data recently to get a current overview.

mutlu_inek · 2009-08-17 12:03:37

Maybe I misread the thread... it seemed to me that it is not intended to be in use right now, but it is quite ambiguous.

So we do another pkgstats drive to get some current data?
Well, do we need them?
Maybe not... it would be good to see how well our last clean-up of [extra] did. And we still have many orphans that we could decide to drop if current usage is small.

gtklocker · 2010-02-16 19:08:33

Whoa, only 7.61 % of Arch users have just installed Konqi !

Shame on the others.

martin77 · 2010-04-10 01:45:24

Chris.Giles wrote:

fukawi2 wrote:
I could see a compromise where pacman has an optional dependancy on pkgstats to help raise awareness of the package among the users (those who use Arch without reading the forums, news etc or who just weren't around when pkgstats was first released)
Yeah, I think having 'pkgstats' as an optional dependency for 'pacman' would be a good minimum. It would make more Arch Linux users aware of 'pkgstats', but there would still be many who'd ignore it and the statistical accuracy would therefore be lower.
If 'pkgstats' becomes a necessary dependency for 'pacman', it could still be optionally disabled in the configuration file or rejected by the user at every hundredth runtime. I don't think this approach would be un-Archie, especially if the 'pkgstats' functionality is disabled by default in the source archive for other distributions.

Hi, I'm been using Arch for some time now and I'm very happy with it. I love it simplicity, it speed, it's cool community, everything, and as hard it may be seen for untrained eyes for install, configure and so on, it excellent aproach to the true meaning of simplicity make Arch a cake to use -with it pros and cons- without sacrificing anything, in fact boosting overall system speed, usability and so on.
Because personal matters -my girlfriend and my shop- I didn't had the time early to check deeply forum/Arch planet for news and what's happening around here (and it twin spanish community, www.archlinux-es.org) but now after dinner I have a little time the first thing I came across was this post about pkgstats.

As fukawi2 says that approach "would make more Arch Linux users aware of 'pkgstats'". Take me as an example. I'm anxious to help back Arch community/devels and if I were aware before about pkgstats I undoubtly would chosed 'in' for it. I don't say Arch will lost a lot because I did not submit my stats, but making it user base aware about this kind of things at least once at 'pacman -Syu' would be great, I think.

However, as dyscoria states, "I must disagree. Sure pkgstats is useful, but to pester the user at all post install (let alone more than once) would be very un-Arch-like, and i'm fairly sure many distributions would share the same view (i know Fedora at least only asks once at install).

The only time a user should ever be asked to provide package statistics is on first install. After installation, providing statistics should be a completely voluntary activity taken by the user. Encouragement to do so on the website or forums is as far as a distribution should go. If an Arch user does not want to provide statistics then he doesn't want to be pestered _at all_ even if it's every thousand times that pacman is run."

I'm curious about how devels with help of the community will find the perfect balance between usability and Arch way. No doubt you're doing an incredible job and because that simple people, un-technical end-users just like me, are starting to be aware about Arch (just take a look at www.distrowatch.org).

Best, and as usual thank you very much for all of you who have helped me pave the way to this great great *great* distro in my box!

martin77 · 2010-04-10 04:19:44

How can I get a complete report of all installed packages in my system as pkgstats do?

If there's not yet that functionality it would be cool to make pkgstats dump the same report it sends to Arch HQ to a desired file so, among other things, I know exactly what will I need in the event I have to install Arch again from scratch - I cross my fingers and touch wood.

* EDIT * I'm pretty sure pckgstats -s > {file} will do the trick

You are awesome guys.

Last edited by martin77 (2010-04-10 04:23:51)

Arch Linux

#101 2009-02-16 09:47:28

Re: pkgstats: contribution was never that easy

#102 2009-02-16 09:55:59

Re: pkgstats: contribution was never that easy

#103 2009-02-16 17:42:34

Re: pkgstats: contribution was never that easy

#104 2009-02-16 21:37:51

Re: pkgstats: contribution was never that easy

#105 2009-02-16 21:47:23

Re: pkgstats: contribution was never that easy

#106 2009-02-17 07:29:18

Re: pkgstats: contribution was never that easy

#107 2009-02-17 20:15:34

Re: pkgstats: contribution was never that easy

#108 2009-02-17 22:12:24

Re: pkgstats: contribution was never that easy

#109 2009-02-18 09:28:29

Re: pkgstats: contribution was never that easy

#110 2009-02-18 09:57:45

Re: pkgstats: contribution was never that easy

#111 2009-02-19 10:06:07

Re: pkgstats: contribution was never that easy

#112 2009-04-16 05:12:55

Re: pkgstats: contribution was never that easy

#113 2009-04-28 18:59:39

Re: pkgstats: contribution was never that easy

#114 2009-05-02 22:21:40

Re: pkgstats: contribution was never that easy

#115 2009-05-08 13:48:41

Re: pkgstats: contribution was never that easy

#116 2009-05-08 22:34:06

Re: pkgstats: contribution was never that easy

#117 2009-05-08 23:01:02

Re: pkgstats: contribution was never that easy

#118 2009-05-08 23:34:41

Re: pkgstats: contribution was never that easy

#119 2009-08-16 17:49:01

Re: pkgstats: contribution was never that easy

#120 2009-08-16 20:10:56

Re: pkgstats: contribution was never that easy

#121 2009-08-16 20:23:05

Re: pkgstats: contribution was never that easy

#122 2009-08-17 12:03:37

Re: pkgstats: contribution was never that easy

#123 2010-02-16 19:08:33

Re: pkgstats: contribution was never that easy

#124 2010-04-10 01:45:24

Re: pkgstats: contribution was never that easy

#125 2010-04-10 04:19:44

Re: pkgstats: contribution was never that easy

Board footer