pkgstats round two: take your vote and help improving Arch

Pierre · 2010-09-25 15:37:39

No, the results are only available for premium users. Otherwise I would have put a link into the first posting, our frontpage or even the pkgstats output.

jarda-wien · 2010-09-25 16:30:29

Pierre wrote:

No, the results are only available for premium users. Otherwise I would have put a link into the first posting, our frontpage or even the pkgstats output.

too bad;-)

marfig · 2010-09-25 17:51:17

I'm not installing. Sorry folks

Unless maybe...

I never quite clearly understood this notion of job prioritization based on incomplete and inconsistent data (as is the case with packages installed on anyone's machine). This is not new to Arch; I've seen it before. But for anyone not using your mainstream package and relying instead on a less common package, this can actually become a problem. At the very least I would hope that this collected data has only so much influence on the developer/maintainer decisions when building their TODO list. As an example, while everyone uses gcc, I bet the fewer that use the sqlite library would rather see the latter given priority over the former.

So, I'm really not sure what is trying to be achieved here. I consider package popularity rather useless data. I'd rather see developers/maintainers using their intuition and their own experience with these packages when doing their TODO list.

Maybe if I could be convinced I'm wrong...

adamlau · 2010-09-25 21:34:43

Stats submitted, any way we can examine our own individual submissions as opposed to the aggregate total?

hokasch · 2010-09-25 22:02:34

Pierre wrote:

Data that are sent by pkgstats:
* list of installed packages without version numbers
* the architecture
* the mirror used (without any username/password scheme)
* the version of pkgstats in use

*pacman -Qq
*uname -m
*cat /etc/pacman.d/mirrorlist
* pacman -Qi pkgstat

apart from that,

Pierre wrote:

1) I don't want track individual users for privacy reasons.

adamlau · 2010-09-25 22:50:49

Heh...Thanks, hokasch . I was too lazy to read past the initial two posts

Allan · 2010-09-25 23:25:48

marfig wrote:

I never quite clearly understood this notion of job prioritization based on incomplete and inconsistent data

It is better than basing it on no information....

At the very least I would hope that this collected data has only so much influence on the developer/maintainer decisions when building their TODO list. As an example, while everyone uses gcc, I bet the fewer that use the sqlite library would rather see the latter given priority over the former.

That is an interesting viewpoint... I think it is highly based on the fact that the packages everyone uses (such as gcc) are already in [core] so are already treated at a higher priority by the developers. It is packages such as those that everyone relies on, but they are taken fore-granted because they rarely break. What is the point of fixing a bug in sqlite if you can not recompile your projects to take advantage of it?

Anyway... last time this informed us that a whole bunch of [extra] getting absolutely no use. A lot of these packages were popular a few years back and had now been superseded (although some were just never popular in the first place...). Using that information, we were able to clean-up the [extra] repo and prioritise our effort towards more popular packages. There is no point spending developer time maintaining or fixing bugs in packages that very few people use - that is what the AUR is for.

At the moment, we have ~400 packages in [extra] that appear not to be installed on any computer. Given the number of submissions, we can be fairly certain they are on less than 1% of Arch computers and should probably be dropped from [extra]. While the numbers are clearly not perfect (due to issues pointed out above), they reflect the installs of people who have contributed towards the information. As always in Arch, if you do not contribute, we do not care about you.

jdarnold · 2010-09-26 00:34:31

But how do you get the IP? My ip is a 192.168 IP address. And I guess you are just getting weekly ideas of the packages out there, so I doubt there's anything about not counting something if it hasn't changed.

RetroX · 2010-09-26 01:15:08

This is actually pretty neat. It would be nice if it tracked AUR/unknown repo packages as well, though.

kazuo · 2010-09-26 01:16:00

Allan wrote:

Anyway... last time this informed us that a whole bunch of [extra] getting absolutely no use. [...] Using that information, we were able to clean-up the [extra] repo and prioritise our effort towards more popular packages.

Agreed, better to have a quality over 9000 packages used by no one. Usually we has a lot of program for doing exactly the same thing and is hard to decide what is the "best" one, the popular one is a good start (at least in Arch I seen to share the taste of the community, the top packages used in the pools are the one I like most.). So having only the most used in the official repos help me to decide what to use in general (sometimes a lost program can be very good).

Allan wrote:

As always in Arch, if you do not contribute, we do not care about you.

One more memorable quote for my quote file!

adamlau · 2010-09-26 01:37:44

Funny quote indeed . Just submitted stats for three ThinkPads .

Snowman · 2010-09-26 02:32:22

RetroX wrote:

It would be nice if it tracked AUR/unknown repo packages as well, though.

It does.

marfig · 2010-09-26 03:22:56

Allan wrote:

Anyway... last time this informed us that a whole bunch of [extra] getting absolutely no use. A lot of these packages were popular a few years back and had now been superseded (although some were just never popular in the first place...). Using that information, we were able to clean-up the [extra] repo and prioritise our effort towards more popular packages. There is no point spending developer time maintaining or fixing bugs in packages that very few people use - that is what the AUR is for.

That's the kind of thing I was hoping for. A convincing argument. Thanks.
Being able to contribute to clean and relevant official repos is definitely something I can do to help you guys.

Allan wrote:

As always in Arch, if you do not contribute, we do not care about you.

But with an hefty price tag at the end, doesn't it? While I appreciate the ulterior motive (make no mistake, you convinced me joining with the above argument) what this essentially means is that participating in this popularity contest you guys put up, has the potential to throw back to Community -- and possibly even AUR -- packages that may be on some user high usability list. Ironically, further shifting the pressure to a repo already low on human resources. "Let someone else handle it", with all the costs that may mean to the end user...

I'm sorry, but I'm not big on quotes when they don't actually reflect the reality. I'd rather have you say on this case "if you don't contribute, you can't help us". Fixed.

EDIT: BTW, it didn't occur anyone putting up a package voting mechanism? Just wondering, because installed packages is just a bad preamble of what actually is going on on every user machine. And the wrong way to describe how exactly users value each package on their machine.

Last edited by marfig (2010-09-26 03:26:40)

mcmillan · 2010-09-26 03:24:48

adamlau wrote:

Stats submitted, any way we can examine our own individual submissions as opposed to the aggregate total?

While hokasch already posted a way to get this, there's also pkgstats -s which will output all the info it's supposed to send

ngoonee · 2010-09-26 09:20:24

marfig wrote:

EDIT: BTW, it didn't occur anyone putting up a package voting mechanism? Just wondering, because installed packages is just a bad preamble of what actually is going on on every user machine. And the wrong way to describe how exactly users value each package on their machine.

Everybody values the packages that they use. Its not an apples-to-oranges comparison when comparing how much anyone 'likes' or 'wants' a particular package. The only 'fair' comparison is the amount of actual usage.

If package foo only has 20 users, it should be in the AUR (unless one of those users is a dev) or community. Even if all 20 users consider it absolutely crucial to their system (if that's the case, they can download the PKGBUILD and makepkg it).

What this does is move the responsibility for package maintainance to those who use the packages, rather than having devs fix bugs on packages with a few dozen users. In a community-driven distro like this, the community needs to take care of all 'specialty' packages.

serprex · 2010-09-26 14:15:36

It'd be best if this only had optional dependencies (Alright, the pacman dependency is a bit hard to factor out), as as of installing this I now have cron installed

Last edited by serprex (2010-09-26 14:15:53)

Pierre · 2010-09-26 16:04:03

Cron is needed to get data regularly. Aside from some special cases, all system should have cron anyway.

ngoonee · 2010-09-27 00:21:45

Out of curiosity, serprex, why don't you have cron installed?

schivmeister · 2010-09-27 07:52:05

jdarnold wrote:

But how do you get the IP? My ip is a 192.168 IP address.

That is your local (private) address, and no-one cares about it. Your router or whatever has the Internet (public) IP.

Snowman wrote:

RetroX wrote:
It would be nice if it tracked AUR/unknown repo packages as well, though.
It does.

Just look at the "unknown" section. Those include AUR and everything else including third-party repos.

birdspider · 2010-09-27 09:48:06

it would be nice to have a link to the packages on the results page
(i.e. "everyone has this package" .. link .. "hmm whats that")

edit: or an ajax tooltip with the pkg description

Last edited by birdspider (2010-09-27 09:50:43)

stryder · 2010-09-27 11:03:54

Just installed. BTW the installation instruction in the news item was "pacman -Sy pkgstats". I thought there was a long discussion on this matter and the conclusion was that the proper way was "pacman -S pkgstats".

DaNiMoTh · 2010-09-27 12:06:49

Just a question: Could be the server side part be installed on others server, without entire archlinux.de environment? We would use this on archppc.

Thanks

Pierre · 2010-09-27 13:32:27

See my mail to arch-dev-public. My plan is to decouple it so its a stand-alone script.

anonymous_user · 2010-09-27 13:43:49

Allan wrote:

Anyway... last time this informed us that a whole bunch of [extra] getting absolutely no use. A lot of these packages were popular a few years back and had now been superseded (although some were just never popular in the first place...). Using that information, we were able to clean-up the [extra] repo and prioritise our effort towards more popular packages. There is no point spending developer time maintaining or fixing bugs in packages that very few people use - that is what the AUR is for.
At the moment, we have ~400 packages in [extra] that appear not to be installed on any computer. Given the number of submissions, we can be fairly certain they are on less than 1% of Arch computers and should probably be dropped from [extra]. While the numbers are clearly not perfect (due to issues pointed out above), they reflect the installs of people who have contributed towards the information.

I'm definitely interested in seeing what results this time.

BTW since installations are a better signal than votes, I wonder if any packages get moved from AUR to community based on their usage.

Last edited by anonymous_user (2010-09-27 13:45:36)

karol · 2010-09-27 13:48:21

There's e.g.

[archstuff]
# AUR's most voted packages
Server = http://archstuff.vs169092.vserver.de/$arch

as a proof that some people care about some other people's votes :-)

Arch Linux

#26 2010-09-25 15:37:39

Re: pkgstats round two: take your vote and help improving Arch

#27 2010-09-25 16:30:29

Re: pkgstats round two: take your vote and help improving Arch

#28 2010-09-25 17:51:17

Re: pkgstats round two: take your vote and help improving Arch

#29 2010-09-25 21:34:43

Re: pkgstats round two: take your vote and help improving Arch

#30 2010-09-25 22:02:34

Re: pkgstats round two: take your vote and help improving Arch

#31 2010-09-25 22:50:49

Re: pkgstats round two: take your vote and help improving Arch

#32 2010-09-25 23:25:48

Re: pkgstats round two: take your vote and help improving Arch

#33 2010-09-26 00:34:31

Re: pkgstats round two: take your vote and help improving Arch

#34 2010-09-26 01:15:08

Re: pkgstats round two: take your vote and help improving Arch

#35 2010-09-26 01:16:00

Re: pkgstats round two: take your vote and help improving Arch

#36 2010-09-26 01:37:44

Re: pkgstats round two: take your vote and help improving Arch

#37 2010-09-26 02:32:22

Re: pkgstats round two: take your vote and help improving Arch

#38 2010-09-26 03:22:56

Re: pkgstats round two: take your vote and help improving Arch

#39 2010-09-26 03:24:48

Re: pkgstats round two: take your vote and help improving Arch

#40 2010-09-26 09:20:24

Re: pkgstats round two: take your vote and help improving Arch

#41 2010-09-26 14:15:36

Re: pkgstats round two: take your vote and help improving Arch

#42 2010-09-26 16:04:03

Re: pkgstats round two: take your vote and help improving Arch

#43 2010-09-27 00:21:45

Re: pkgstats round two: take your vote and help improving Arch

#44 2010-09-27 07:52:05

Re: pkgstats round two: take your vote and help improving Arch

#45 2010-09-27 09:48:06

Re: pkgstats round two: take your vote and help improving Arch

#46 2010-09-27 11:03:54

Re: pkgstats round two: take your vote and help improving Arch

#47 2010-09-27 12:06:49

Re: pkgstats round two: take your vote and help improving Arch

#48 2010-09-27 13:32:27

Re: pkgstats round two: take your vote and help improving Arch

#49 2010-09-27 13:43:49

Re: pkgstats round two: take your vote and help improving Arch

#50 2010-09-27 13:48:21

Re: pkgstats round two: take your vote and help improving Arch

Board footer