As promised, a more detailed reply. (Sorry it took me a bit—my laptop was crashing strangely.)

Arjan's main question was about monitoring vandalism across multiple incubator wikis, which are suggested in the proposal. And my answer is no—at the moment I don't have more details about it.

Obviously, this cannot be done without a clear and agreed solution to this problem, or at least an agreement to make a temporary conditional experiment. That's why I wrote in the Phabricator task: "Patroling all such domains for vandalism must be at least as easy as it is for Incubator". I don't have a complete solution to this, but it's clear that it's a sine-qua-non condition for moving forward.

My plan was to bring this up at Incubator's own community discussion pages.

The closest thing that I have currently to a written proposal to monitor wikis is the one by Martin Urbanec at his comment on Phabricator:
https://phabricator.wikimedia.org/T165585#4221392

Does this sound like a beginning of something sensible and feasible? Or is it too technical and complex and inappropriate?


--
Amir Elisha Aharoni · אָמִיר אֱלִישָׁע אַהֲרוֹנִי
http://aharoni.wordpress.com
‪“We're living in pieces,
I want to live in peace.” – T. Moore‬

2018-07-26 11:32 GMT+03:00 Arjan Verheijden <arjanverheijden@hotmail.com>:

Hi!


I have to tell that I am not very happy about this being discussed without consulting the admins of the Incubator. There has been no announcement about this discussion, nor even a simple message on admin's talk pages about this. If I weren't subscribed to the LangCom mailing list, I wouldn't even have known this discussion was taking place.


As a lot is pretty technical, I can only comment on what I think this proposal is about.


I agree that there are some major issues on the Incubator concerning edit comfort (prefixes, standard templates/modules, wikidata, etc.). If I understand correctly, the idea is to split off the most active tests into own semi-subdomains. A big problem that will have to be addressed with this is that separate domains require way more monitoring. Now, with all tests in one wiki, it is easy for the Incubator team to remove spam, vandalism, errors in pages, etc, and easy to run bots for broken redirects etc. With separate domains, that will become an impossible task, unless somekind of admin interface is created in which all tests can be monitored from one Recent Changes.


I saw in the Phabricator task: "Patroling all such domains for vandalism must be at least as easy as it is for Incubator." Is anything known about the status of this? Has a solution already been found?


Greetings,

Owtb




Från: Langcom <langcom-bounces@lists.wikimedia.org> för Amir E. Aharoni <amir.aharoni@mail.huji.ac.il>
Skickat: den 26 juli 2018 07:42
Till: Wikimedia Foundation Language Committee
Ämne: Re: [Langcom] Please read: Fixing Incubator
 


2018-07-25 20:04 GMT+03:00 Steven White <Koala19890@hotmail.com>:
>... I really don't want this
>discussion to go in the direction of arguing about historical languages
>policy. ...

Me, either. But for the moment Incubator needs to continue to serve those test projects.

No problem.


>... I even welcome you to create semi-artificial
>criteria that would put the number of the first test wikis at around 50 :)

I can manage that.

>I guess that I'd do one of the following:
>
>1. Define what does "a while" mean. For example, we can decide, somewhat
>arbitrarily, that a project under incubator.wikimedia.org in which there
>were no edits for over a year, will be considered "dormant" and that a test
>wiki won't be created for it until somebody asks.
>2. Just use case by case intuition.

Right now I am using a blend of the above. I tend to use page creations, not edits, because from time to time you get random edits from people who sweep through and update the name of the current prime minister in the infobox, without really doing anything else, and I'm not sure those really reflect "activity".

They don't. It should all be done in Wikidata, for a lot of reasons, and this is one of those reasons.
 
That said, I also ignore page additions by certain people I know are doing maintenance (me, Liuxinyu970226, others). I also ignore cases where the only page addition comes from the automatic creation of a redirect after a page move. Finally, for this purpose, we ignore single page creations from IPs. 

>>       - Conversely, many tests open with a flurry of activity (over 1 day
>>       to 2 months), then go dormant.
>
>So, here's my theory behind the whole proposal: In the current incubator,
>people quickly create a bunch of articles on topics that interest them,
>sometimes with some boilerplate (cities, countries, animals), but but then
>they get tired of the prefixes, the outdated translation techniques, the
>weird searching, the missing templates, etc., and give up. Perhaps with a
>single wiki these difficulties will be alleviated. I know it sounds a bit
>too optimistic, but at least I want to get rid of these most glaring and
>arbitrary difficulties, with the hope that it will help people remain
>active.

You may be right. I just don't know. Do we have feedback saying so? Or is this just everyone's gut instinct. (Mind you, I can't possibly say I disagree.) 

Observing people in real life editing in Incubator, many times. In particular, in Kabardian, Adyghe, Fon, and Dinka, with which I've had a closer relationship.

And a few days ago there was a presentation at Wikimania that makes pretty much the same complaints: https://www.youtube.com/watch?v=yMOS19Dj7rU&t=29m .

 

That said, this approach doesn't solve all the problems you describe. In particular, this won't automatically create basic templates and modules,

Yes. I'm working on that in a separate initiative. It's complicated, but much-needed.
 
And at the end of the day, contributors to new tests still have to write or translate pages.

Of course, as they do in usual wikis. It's not without issues, but at least it will have proper search, Wikidata, and Content Translation, and no prefixes.
 

>Yes, I tend to think that we should allow to create new subdomains for
>*eligible* languages right away, so that all of them will get equal
>treatment as early as possible, from the very first page.

In order to do this, then, a few things would have to happen in fairly short order, once we're ready to approach the problem like that.

  • The "Requests for new languages" system on Meta will have to have teeth, and to be the arbiter for whether a project is (a) eligible for an incubation subdomain, (b)(interim, at least) should start on Incubator, or (c) is not eligible. IF THAT IS TO BE MEANINGFUL, I NEED LANGCOM MEMBERS TO RESPOND TO MY SUBMISSIONS ON SUCH MATTERS, or to go to Meta and mark projects eligible themselves. I appreciate that people are generally allowing me to call those shots right now, but since just about anyone can start a test project on Incubator, the stakes on decisions about "eligible" aren't all that high right now. The stakes will become higher under this new regimen.

Yes, this is pretty obvious.

I do envision that the incubator wiki creation will a mostly automatic process. There is no good reason for it to require any manual intervention from Ops engineer, as it happens now. There should be a simple form somewhere on one of our other wikis, probably Meta or wikitech.wikimedia.org , where one will have to fill project family (pedia, voyage, etc.), language code, autonym, text direction, logo file name—and push a button. The interesting question is who should have the permission to fill this form and push the button. My immediate thought is "language committee members", of course, but other ideas are welcome.


  • FWIW, in clearing the backlog at RFL, we're almost to the end of requests dating to 2012 or earlier, and have addressed almost all of the 2017 and 2018 requests (except those in the last month or so).

Have I mentioned how much do I appreciate what you do there? It's truly wonderful that you are taking care of this backlog.


_______________________________________________
Langcom mailing list
Langcom@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/langcom