You are not logged in.

#26 2005-08-24 04:24:16

chane
Member
Registered: 2003-12-02
Posts: 93

Re: Why the AUR voting system is not very popular

arooaroo wrote:

Well then it would probably be very obvious then that there's an unnatural surge in activity. Of course, I never envisiged a system as crude as raw hit counts. But at the end of the day, how ruthless do you think Archers are?

I'm going to go out on a limb and ask why not count raw hits/downloads.  Sure people could artifically bump up a specific package, but who is going to want to do that and if they do, how many different version of the same package are they going to do it for?  Also, you could watch the hit counts for a few versions of the package.  You might have an initial spike when a package is first released, however, the real interesting part should be how many people download it when there is a new version (I'm assuming this would mainly occur with pacman -Syu).

Then there is the issue of someone like myself who internally mirrors the repos (only the official ones), so at most you'll get 1 hit from me using rsync on everything.  I run pacman from my internal repos for all of the arch packages.  I personally will not install ArchStats on any of my servers.  We run as few things as needed on those.  And I don't run any desktops to speak of.  However, I am probably in the minority for having an internal repo (we do it for testing purposes and to have a consistent snapshot for all of our machines).

Just a comment from someone who really likes arch (and pacman) and lurks around a bit.

Chris....

Offline

#27 2005-08-24 09:03:54

dtw
Forum Fellow
From: UK
Registered: 2004-08-03
Posts: 4,436
Website

Re: Why the AUR voting system is not very popular

This is not an effective solution for one reason downloads do not equal popularity e.g. 500 people donwload pkg A but it's useless so 450 remove it straight away, but the TUs see 500 downloads and it goes straight into [community]

People could just vote - if people have time to use the AUR they should have time to vote...

Offline

#28 2005-08-24 13:01:04

chane
Member
Registered: 2003-12-02
Posts: 93

Re: Why the AUR voting system is not very popular

dibblethewrecker wrote:

This is not an effective solution for one reason downloads do not equal popularity e.g. 500 people donwload pkg A but it's useless so 450 remove it straight away, but the TUs see 500 downloads and it goes straight into [community]

If you read what I wrote, I said you should probably monitor the downloads over a few versions.  Yes people do download and then remove.  That's why it would be important to trend it over a couple of releases.

dibblethewrecker wrote:

People could just vote - if people have time to use the AUR they should have time to vote...

That's the entire point.  People are not voting.  Good luck appealing to people to do this.  I actually myself do vote.  However, I know human nature well enough to know this will not happen to the vast degree you want it to, particularly as the distro grows (or in the main repos that are easily gotten via pacman).

I realized you are focused on the AUR.  It would be nice if a scheme could be devised that also could be applied to the official repos.  That way the devs could get an idea of how there stuff is being used.

Anyway, it was just a comment, good luck and I hope you come up with a great solution.

Chris....

Offline

#29 2005-08-24 13:47:02

arooaroo
Member
From: London, UK
Registered: 2005-01-13
Posts: 1,268
Website

Re: Why the AUR voting system is not very popular

dibblethewrecker wrote:

This is not an effective solution for one reason downloads do not equal popularity e.g. 500 people donwload pkg A but it's useless so 450 remove it straight away, but the TUs see 500 downloads and it goes straight into [community]

People could just vote - if people have time to use the AUR they should have time to vote...

Of course, I voted for the ati-drivers and the cko kernel, and even though I haven't actually removed them, both are now collecting dust as I've gone back to stock kernels and drivers for now.

So, I supose votes may not truely reflect usage either. Having said that, I think it's the best way and is far less prone to error than just download stats alone.

Offline

#30 2005-08-24 22:31:01

dtw
Forum Fellow
From: UK
Registered: 2004-08-03
Posts: 4,436
Website

Re: Why the AUR voting system is not very popular

chane wrote:
dibblethewrecker wrote:

This is not an effective solution for one reason downloads do not equal popularity e.g. 500 people donwload pkg A but it's useless so 450 remove it straight away, but the TUs see 500 downloads and it goes straight into [community]

If you read what I wrote, I said you should probably monitor the downloads over a few versions.  Yes people do download and then remove.  That's why it would be important to trend it over a couple of releases.

And if you read what i wrote i don't see how a download trend over an infinite number of release will ever reflect popularity.

When I sort downloads on websites I sort by rating, not by downloads, downloads to me say "It looks good to lots of people" that doesn't mean it is good.

I personally think, given some serious time, the reason people rarely vote for stuff in the AUR is because 90% of stuff is not worth voting for - 25 people have not even used it, never mind like it.  I think we need to accept that pkgs will rarely be added to [community] because there just aren't enough good pkgs out there.

Either that or we need to reduce the number of votes necessary to add a pkg to [community]

Offline

#31 2005-08-24 22:51:20

arooaroo
Member
From: London, UK
Registered: 2005-01-13
Posts: 1,268
Website

Re: Why the AUR voting system is not very popular

dibblethewrecker wrote:

Either that or we need to reduce the number of votes necessary to add a pkg to [community]

I think 25 is reasonable. I would be against reducing it for now. As AUR becomes more familiar, and as the AL userbase grows, I'm sure 25 will soon feel like a very low threshold.

Offline

#32 2005-08-24 23:36:30

Snowman
Developer/Forum Fellow
From: Montreal, Canada
Registered: 2004-08-20
Posts: 5,212

Re: Why the AUR voting system is not very popular

One reason that some AUR packages have low votes is that they are very specialized.  Only a few persons needs/uses them.

25 is a good threshold.  Also, there is nothing preventing a TU to add packages that have received less than 25 votes to the [community] repo.

Offline

Board footer

Powered by FluxBB