FS#45326 - Suggestion: Show unique IP count of downloads from last X months, instead of 'Popularity'
Attached to Project:
AUR web interface
Opened by (Det) - Sunday, 14 June 2015, 12:23 GMT
Last edited by Lukas Fleischer (lfleischer) - Sunday, 14 June 2015, 12:37 GMT
Opened by (Det) - Sunday, 14 June 2015, 12:23 GMT
Last edited by Lukas Fleischer (lfleischer) - Sunday, 14 June 2015, 12:37 GMT
|
Details
I would imagine showing, e.g. the unique IP count of
downloads, e.g. the last month, 3 months or so would be a
more relevant metric than the "Popularity" score (which I
haven't seen explained anyhwere besides the commit[1] and
[aur-dev][2] (like the home page[3] or just hovering
over)).
This would mean that both disadvantages in counting downloads, meaning 1) older packages getting a hude lead, and 2) packages getting updated more often get more downloads, aren't counted. Dynamic IPs, I don't think, affect it that much. In Popularity, if I understand it, after a hundred days, your vote will only be worth: 0.98 ^ 100 ≈ 0.13 (~87% less). Well. You might just as well still use the thing and find it more lovely than ever, this seems too much like having old, unused user accounts dragging the significance of the entire active user base down with them (which assumably is the majority anyway). The removed accounts' votes back in February[4] were also discarded from the total votes. [1] = https://projects.archlinux.org/aurweb.git/commit/?id=824744f1d20614c25e972dda7a0b7ac9506dd46f [2] = https://lists.archlinux.org/pipermail/aur-dev/2015-June/003254.html [3] = https://aur4.archlinux.org [4] = https://lists.archlinux.org/pipermail/aur-general/2015-February/030231.html |
This task depends upon
Closed by Lukas Fleischer (lfleischer)
Sunday, 14 June 2015, 12:37 GMT
Reason for closing: Won't implement
Sunday, 14 June 2015, 12:37 GMT
Reason for closing: Won't implement
Comment by
Lukas Fleischer (lfleischer) -
Sunday, 14 June 2015, 12:37 GMT
At least where I live, we use dynamic IP addresses and your public
IP address changes every day automatically. This means that
packages getting updated more often definitely get more hits (a
package getting updated every day gets ~30 times the number of
hits of a package that is never updated if we only consider one
month). I also don't understand how it fixes any problem with the
current measure: If there is a lovely tool you voted for ~1 year
ago, your IP address also won't show up in the list of last
month's downloaders (unless the package is updated frequently).
Furthermore, it seems completely unlikely that there are 100
people who downloaded a tool months ago and no new users at all.
And even if that really is the case, it means that the tool seems
to be of no interest for new users which should be reflected in
the popularity score.