Don't assign karma for published translation uploads

Reported by Данило Шеган on 2009-03-03
24
Affects Status Importance Assigned to Milestone
Launchpad itself
Medium
Данило Шеган

Bug Description

For "published" uploads (i.e. coming from Soyuz or maintainer uploads), we should not assign any karma.

Changed in rosetta:
importance: Undecided → Medium
status: New → Triaged
Karl Fogel (kfogel) wrote :

I *think* this might be a dup of bug #306762. Not positive; mentioning it here so someone more qualified can determine.

Sort of: that one is actually about removing karma which was assigned due to this problem when only a few people could cause that (like Jeroen, me, and a few other people who are/were members of rosetta-admins). Of course, they are very closely related, and it'd be worth removing karma for people who got their karma due to this bug recently as well.

We are giving too much karma for template and upstream translation file import: 10 and 30 points respectively.

We should make that more along the lines of 2 and 1 (or similar). Still, for people who upload many packages, this might turn out to be a lot.

For instance, Colin Watson has uploaded around 5500 translation files in the last month, and 123 templates. By the new calculation that would give him ~5800 karma points. Also, a KDE packager Steve Stalcup (vorian) has uploaded 22500 translation files and 750 templates, which would amount to ~24000 karma points over the span.

Though, Steve is the one doing most uploads, and 22500 translations is across ~100 KDE language pack uploads, so maybe he deserves the karma?

I'd need to check with our karma gurus what they think is sane.

Note that Steve today has over 1.3M karma points in Translations (a sum of karma points would be around 650k, so I assume it's being doubled for recent contributions). Sebastian Bacher is the other "large" uploader (GNOME packager) who has uploaded around 11000 translation files.

How to best scale this to compare with:
 ∙ providing a suggestion, 3 points
 ∙ reviewing a suggestion, 3 points
 ∙ doing a translation (i.e. you are a reviewer), 3 points

I.e. if we assign a single point for a translation upload someone would have to do around 4000 translations to be comparable with Seb, or 7000 to be comparable to Steve. Still, we should note that these two guys are "worst" cases:
 vorian | 22426
 seb128 | 11344
 mvo | 6303
 cjwatson | 5553
 mdke | 5002
 doko | 2024
 pitti | 1970
 didrocks | 1807
 arnegoetje | 1761
 lure | 1597
 ember | 1579
 andreas-moog | 1422
 kees | 1120
 asac | 1090
 chrisccoulson | 930
 robert-ancell | 911
 ken-vandine | 766
 themuso | 718
 ccheney | 680
 jdstrand | 660

As an alternative, we can just stop giving out karma for these uploads except for templates. In general, I feel that's the right choice.

Changed in rosetta:
assignee: nobody → danilo
milestone: none → 2.2.4
Matthew East (mdke) wrote :

If you reduce the number of karma points for translations, that will be unfair on real translators.

The right solution is definitely to stop giving karma for package uploaders, because (a) they now get karma for the soyuz upload, and (b) they haven't done the translation themselves.

Matthew, this won't affect any translations done by users directly (i.e. either through an upload or via web UI). We assign these only for file uploads which are marked as "published" (from maintainers, soyuz, etc.).

It's just that for every single published upload we give out too much karma, when we might not need to give any at all. I'd still like to give out some karma for template uploads, because that means that people are using translations even if they are not translating (i.e. for people hosting their projects and translations in LP).

So, I'd go with 1 point for POT imports, 0 for PO imports.

Matthew East (mdke) wrote :

It's certainly fair to grant karma for POT imports and not for PO imports when they come through Soyuz. As to whether 1 point is appropriate, I'll leave that to your karma guys, but seeing that you get 5000 points for a package upload, 1 point sounds disproportionately small.

The thing is that in majority of cases, packagers don't really bother with translations at all, and just deliver what is in the upstream tarball after a POT is generated. For instance, Ubuntu has several packagers which upload a lot of packages, and they get their package upload karma (which is actually 10 or 5 or 1, depending on the type, and not 5000) and POT upload karma (1-10 points, depending on the number of packages uploaded).

Some people provide hundreds of POTs every few weeks through package uploads, which means that for a translator to be comparable to them, he'd have to provide same number of actual translations. I believe that's disproportionate as well.

Of course, this would be unfair to people (like you) who do their "published" imports manually, but at the moment, there is no way for us to make a distinction, so we can't fix that in an easy way.

On Wed, Apr 8, 2009 at 11:52 AM, Данило Шеган <email address hidden> wrote:
> The thing is that in majority of cases, packagers don't really bother
> with translations at all, and just deliver what is in the upstream
> tarball after a POT is generated. For instance, Ubuntu has several
> packagers which upload a lot of packages, and they get their package
> upload karma (which is actually 10 or 5 or 1, depending on the type, and
> not 5000) and POT upload karma (1-10 points, depending on the number of
> packages uploaded).

Ok, apologies if I've got the 5000 points wrong - I reviewed my own
karma, and the Soyuz karma appears to have gone up in blocks of 5000 -
https://launchpad.net/~mdke/+karma

> Some people provide hundreds of POTs every few weeks through package
> uploads, which means that for a translator to be comparable to them,
> he'd have to provide same number of actual translations. I believe
> that's disproportionate as well.

Agreed. Ideally the karma system would recognise what has actually
changed in an uploaded pot template and limit the karma assigned to
those strings.

> Of course, this would be unfair to people (like you) who do their
> "published" imports manually, but at the moment, there is no way for us
> to make a distinction, so we can't fix that in an easy way.

Well, we import through package uploads too, so no problem there.

Changed in rosetta:
status: Triaged → In Progress

This has now been fixed on production, I just need to land the branch with sampledata updates.

Sampledata fixed in devel 8224, and the updated karma values are out in the wild.

Changed in rosetta:
status: In Progress → Fix Released
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Duplicates of this bug

Other bug subscribers