Hello,
is it possible to see the number of downloads of packages from AUR, so that it can be detected, how much interest in a package exists?
Say, there are some users who do not have a AUR-login, and just would install the packages that are there... which are possibly outdated, but would nevertheless be interested in installing newer package, if possible.
If the download number is high enough, even if there are not much votes (because some people may only install stuff but are not interested in package maintaining and so on), then at least interest of a package might be detected this way.
I'm asking, because I think I can adopt some more packages, but would of course only pick those that I find interesting. They might be interesting (topic), even if I wil not use them by myself. So, from that standpoint, I would then select by my own interest and that of other people. For packages that I use by myself, of course my interest is clear, and I would pick such packages.
But say there are 100 orphaned packages and 20 of them look interesting, but I would not use them by myself, then I would adopt those, who are wanted by many people. (And any package that I want to install myself.)
I hope I could make ckear, why I'm asking.
Say, there are 5 interesting packages, but 3 of them will not be used, because that stuff is not used any more (maybe because better projects are to prefer), then three don't need support I think.
So, the tarball download numbers do matter in this respect.
Is it possible to add them to the AUR-pages of a project?
Ciao, Oliver
On Thu, Feb 28, 2013 at 09:09:30PM +0100, oliver wrote:
Hello,
is it possible to see the number of downloads of packages from AUR, so that it can be detected, how much interest in a package exists?
Say, there are some users who do not have a AUR-login, and just would install the packages that are there... which are possibly outdated, but would nevertheless be interested in installing newer package, if possible.
If the download number is high enough, even if there are not much votes (because some people may only install stuff but are not interested in package maintaining and so on), then at least interest of a package might be detected this way.
This is an extremely flawed premise.
I'm asking, because I think I can adopt some more packages, but would of course only pick those that I find interesting.
They might be interesting (topic), even if I wil not use them by myself. So, from that standpoint, I would then select by my own interest and that of other people. For packages that I use by myself, of course my interest is clear, and I would pick such packages.
This is really strange (non-)logic. The best maintainer for a package is one who is actively interested in the package itself and uses it. How else could you possibly support it?
But say there are 100 orphaned packages and 20 of them look interesting, but I would not use them by myself, then I would adopt those, who are wanted by many people. (And any package that I want to install myself.)
I hope I could make ckear, why I'm asking.
Say, there are 5 interesting packages, but 3 of them will not be used, because that stuff is not used any more (maybe because better projects are to prefer), then three don't need support I think.
So, the tarball download numbers do matter in this respect.
I disagree. Downloading a tarball doesn't mean you intend to build, install, and use it.
Is it possible to add them to the AUR-pages of a project?
And display this how? A flat "download count" would be of little interest or value. Would it be reset every time there's a new PKGBUILD uploaded? Maybe only for pkgver changes? Would that include pkgrel bumps too? What about historical data? Trending? Unique by IP? Unique by IP within a given date range?
If you want to make download count a useful metric, you have to do a lot more than display a raw counter. As you propose it, it's no better (and probably worse) than a vote count.
On Thu, Feb 28, 2013 at 03:19:09PM -0500, Dave Reisner wrote:
On Thu, Feb 28, 2013 at 09:09:30PM +0100, oliver wrote:
Hello,
is it possible to see the number of downloads of packages from AUR, so that it can be detected, how much interest in a package exists?
Say, there are some users who do not have a AUR-login, and just would install the packages that are there... which are possibly outdated, but would nevertheless be interested in installing newer package, if possible.
If the download number is high enough, even if there are not much votes (because some people may only install stuff but are not interested in package maintaining and so on), then at least interest of a package might be detected this way.
This is an extremely flawed premise.
I'm asking, because I think I can adopt some more packages, but would of course only pick those that I find interesting.
They might be interesting (topic), even if I wil not use them by myself. So, from that standpoint, I would then select by my own interest and that of other people. For packages that I use by myself, of course my interest is clear, and I would pick such packages.
This is really strange (non-)logic. The best maintainer for a package is one who is actively interested in the package itself and uses it. How else could you possibly support it?
Of course, as I said, I would NOT adopt packages that are completely uninteresting for me. Nevertheless, if I think a topic is interesting, and is somehow related to something I would like to see spreaded, adopting it can make sense.
For example, if I want to push the usage of R, and there are GUIs for R, that many people would like to use, supporting a GUI for R can mek sense, even I prefer the shell without GUI for myself. But if it contributes to spreading R - which I think is a wonderful program - it would make sense. E.g. I could show the GUI to some friends, who are GUI-addicted, and in this way introduce them to R.
So I would not use it, but would like to provide it. But it makes no sense, if nobody asks for it.
Do you see what I have in mind?
Do you nevertheless think, this is weak motivation?
About the metrics: download numbers are not perfect, but better than nothing?
Ciao, Oliver
On Thu, Feb 28, 2013 at 09:34:11PM +0100, oliver wrote:
On Thu, Feb 28, 2013 at 03:19:09PM -0500, Dave Reisner wrote:
On Thu, Feb 28, 2013 at 09:09:30PM +0100, oliver wrote:
Hello,
is it possible to see the number of downloads of packages from AUR, so that it can be detected, how much interest in a package exists?
Say, there are some users who do not have a AUR-login, and just would install the packages that are there... which are possibly outdated, but would nevertheless be interested in installing newer package, if possible.
If the download number is high enough, even if there are not much votes (because some people may only install stuff but are not interested in package maintaining and so on), then at least interest of a package might be detected this way.
This is an extremely flawed premise.
I'm asking, because I think I can adopt some more packages, but would of course only pick those that I find interesting.
They might be interesting (topic), even if I wil not use them by myself. So, from that standpoint, I would then select by my own interest and that of other people. For packages that I use by myself, of course my interest is clear, and I would pick such packages.
This is really strange (non-)logic. The best maintainer for a package is one who is actively interested in the package itself and uses it. How else could you possibly support it?
Of course, as I said, I would NOT adopt packages that are completely uninteresting for me. Nevertheless, if I think a topic is interesting, and is somehow related to something I would like to see spreaded, adopting it can make sense.
For example, if I want to push the usage of R, and there are GUIs for R, that many people would like to use, supporting a GUI for R can mek sense, even I prefer the shell without GUI for myself. But if it contributes to spreading R - which I think is a wonderful program - it would make sense. E.g. I could show the GUI to some friends, who are GUI-addicted, and in this way introduce them to R.
So I would not use it, but would like to provide it. But it makes no sense, if nobody asks for it.
Do you see what I have in mind?
No. I don't.
Do you nevertheless think, this is weak motivation?
Yes, I do. Adopt things because you're interested in them, not because they're (potentially?) popular.
About the metrics: download numbers are not perfect, but better than nothing?
What you propose is surfacing raw *data*. I made a counter proposal of adding potential heuristics so that useful *information* could be displayed. I do not believe there is any value in showing the raw data, and, if anything, I believe it would only be misleading.
d
On Thu, Feb 28, 2013 at 03:39:04PM -0500, Dave Reisner wrote:
On Thu, Feb 28, 2013 at 09:34:11PM +0100, oliver wrote:
On Thu, Feb 28, 2013 at 03:19:09PM -0500, Dave Reisner wrote:
On Thu, Feb 28, 2013 at 09:09:30PM +0100, oliver wrote:
Hello,
is it possible to see the number of downloads of packages from AUR, so that it can be detected, how much interest in a package exists?
Say, there are some users who do not have a AUR-login, and just would install the packages that are there... which are possibly outdated, but would nevertheless be interested in installing newer package, if possible.
If the download number is high enough, even if there are not much votes (because some people may only install stuff but are not interested in package maintaining and so on), then at least interest of a package might be detected this way.
This is an extremely flawed premise.
I'm asking, because I think I can adopt some more packages, but would of course only pick those that I find interesting.
They might be interesting (topic), even if I wil not use them by myself. So, from that standpoint, I would then select by my own interest and that of other people. For packages that I use by myself, of course my interest is clear, and I would pick such packages.
This is really strange (non-)logic. The best maintainer for a package is one who is actively interested in the package itself and uses it. How else could you possibly support it?
Of course, as I said, I would NOT adopt packages that are completely uninteresting for me. Nevertheless, if I think a topic is interesting, and is somehow related to something I would like to see spreaded, adopting it can make sense.
For example, if I want to push the usage of R, and there are GUIs for R, that many people would like to use, supporting a GUI for R can mek sense, even I prefer the shell without GUI for myself. But if it contributes to spreading R - which I think is a wonderful program - it would make sense. E.g. I could show the GUI to some friends, who are GUI-addicted, and in this way introduce them to R.
So I would not use it, but would like to provide it. But it makes no sense, if nobody asks for it.
Do you see what I have in mind?
No. I don't.
Do you nevertheless think, this is weak motivation?
Yes, I do. Adopt things because you're interested in them, not because they're (potentially?) popular.
Yes.
Why I was looking for poularity? Because I thought, it *might* be an indicator (even a weak one) about some kind of quality. Some libraries or tools might not be used anymore, because there are newer or faster or better libs for example.
Maybe it would be better, asking people about quality of a software, instead of just looking at the numbers... hmhhh. Maybe the numbers are really to weak to decide this.
But somehow I think, if something is used often, there might be a bigger number of users who can support other people. Hence more prgress, and (hopefully) rising quality.
Often people say: a project that was not updates since years is a dead project. People turn away. Isn't this most often (of course not always, not hgenerally) some kind of indicator to see some kind of quality behind these decisions?
About the metrics: download numbers are not perfect, but better than nothing?
What you propose is surfacing raw *data*. I made a counter proposal of adding potential heuristics so that useful *information* could be displayed. I do not believe there is any value in showing the raw data, and, if anything, I believe it would only be misleading.
Ok, I see.
Ciao, Oliver
keyword that I had in mind: swarm intelligence
On Thu, Feb 28, 2013 at 10:02 PM, oliver oliver@first.in-berlin.de wrote: <snip>
Yes.
Why I was looking for poularity? Because I thought, it *might* be an indicator (even a weak one) about some kind of quality. Some libraries or tools might not be used anymore, because there are newer or faster or better libs for example.
<snip>
Welcome on the planet where linux is being used because it comes with apache, php and mysql. Quality is an elastic word, it's not necessary to stress it further with some shady means of heuristically aur packages' popularity. why is everyone thinking they are rockstars because they have packages on aur? it's work you do because it's fun to you, and so is all work.
Please, let's remove the voting function from the aur alltogether (have I ever voted?) and let's not run any further into the "package wars". there are perfectly fine reasons why two packages that basically do the same thing might not be equally popular. One might be built for an edge case the other just barely doesn't cover. that's how nature works in a nondestructive environment, and that's why the aur is awesome as the useful zoo it is.
cheers! mar77i
On 2013-02-28 at 21:09:30 +0100, oliver wrote:
is it possible to see the number of downloads of packages from AUR, so that it can be detected, how much interest in a package exists?
I suspect that the use case of "download from AUR, try to build and install, (possibly even fail, or just decide that another alternative fits better), delete the package, forget about it" is not so rare, expecially with old orphaned packages.
The system used in Debian to estimate the popularity of packages, popcon_, gives the number of installed packages, but also what IMHO is an even more important stats: the number of recently used installation.
This requires voluntary submissions, the installation of an ad-hoc program and working atime on /usr, so it is not perfect and probably not something that arch can adopt, but IMHO worth considering as a comparison.
.. _popcon: http://popcon.debian.org/
On Fri, Mar 01, 2013 at 09:47:28AM +0100, Elena ``of Valhalla'' wrote:
On 2013-02-28 at 21:09:30 +0100, oliver wrote:
is it possible to see the number of downloads of packages from AUR, so that it can be detected, how much interest in a package exists?
I suspect that the use case of "download from AUR, try to build and install, (possibly even fail, or just decide that another alternative fits better), delete the package, forget about it" is not so rare, expecially with old orphaned packages.
Does "orphaned" mean old, orphaned packages? I saw some entries, where maintenance was stopped because of "not time". If such a package has 6 votes and 200 hundred downloads, it looks more accepted/used as one with 0 votes and 7 downloads.
If both cover interesting topic (maybe the same topic), it's not guaranteed, but maybe more likely that the one with more votes and downloads might need a maintainer.
The system used in Debian to estimate the popularity of packages, popcon_, gives the number of installed packages, but also what IMHO is an even more important stats: the number of recently used installation.
Aha. But is that really more accurate? A new installation could mean someone who is installing Linux the first time, and then tries some packages... ...is there also a statistics about package removal?
This requires voluntary submissions, the installation of an ad-hoc program and working atime on /usr, so it is not perfect and probably not something that arch can adopt, but IMHO worth considering as a comparison.
.. _popcon: http://popcon.debian.org/
-- Elena ``of Valhalla''
Thanks you.
Ciao, Oliver
On 2013-03-01 at 13:44:04 +0100, oliver wrote:
Aha. But is that really more accurate? A new installation could mean someone who is installing Linux the first time, and then tries some packages... ...is there also a statistics about package removal?
err, recently *used*, as in "the atime on the binary is in the last X days". It's not perfect, it does not count libraries and it does fail if atime is not available, but usually it works.
recent installs/updates are counted separately.
On Fri, Mar 01, 2013 at 03:03:18PM +0100, Elena ``of Valhalla'' wrote:
On 2013-03-01 at 13:44:04 +0100, oliver wrote:
Aha. But is that really more accurate? A new installation could mean someone who is installing Linux the first time, and then tries some packages... ...is there also a statistics about package removal?
err, recently *used*, as in "the atime on the binary is in the last X days". It's not perfect, it does not count libraries and it does fail if atime is not available, but usually it works.
[...]
Aha, nice. :-)
Thanks for explanation.
Ciao, Oliver
aur-general@lists.archlinux.org