Talk:Requests for comment/Opt-in site registration during installation
Wiki description Edit
This information would be good to have in core, and be output in <meta name="description" /> on the main page. That would make Google and Bing display that description when the main page is on search results (for example, searching for the name of the wiki). It should be a small sentence, though, different from a hypothetical Project:About (mediawiki:aboutpage). This was done on Wikia years ago, where wikis can edit that on MediaWiki:Description. --Ciencia Al Poder (talk) 09:45, 24 September 2013 (UTC)
Data sent Edit
I don't understand the proposed data. Why send data that we can easily retrieve via the API? In essence, the only thing we need is the API endpoint URL (or index.php for older wikis and wikis where it's disabled, but those are probably both hopeless), plus some other unstructured data not in the API (as a description, and for many wikis the copyright status) and perhaps some server environment information (mentioned e.g. in mailarchive:wikitech-l/2013-October/072185.html). --Nemo 00:04, 2 October 2013 (UTC)
- The reason to offer to send the data to a central collection point is so that we can find out about new wikis without having to discover them. This would also allow us to collect information on wikis that are not publically accessible if the wiki creators allow it. — ☠MarkAHershberger☢(talk)☣ 02:43, 24 April 2014 (UTC)
- It still feels unnecessarily complex: we don't need most of this stuff. All in all, providing a link to https://wikiapiary.com/wiki/Special:FormEdit/Website next to the "subscribe to mediawiki-announce" checkbox might be enough. I think we've never even announced the existence of WikiApiary on mediawiki-announce!
- I mean, all this is nice and I would be very happy if we had a perfect solution, but it has an infinite potential for bikeshedding and politicking. If there is interest in the general goal "let's find out who's using MediaWiki", I'd really love us to focus on the low-hanging fruit. One user who got 22k wikis added to WikiApiary 10:08, 3 December 2014 (UTC)
Publicity of the data sent Edit
Generally, I think we need this data desperately in order to improve MediaWiki for 3rd parties. Just a few remarks:
- Some (if not many) of the wikis will be inhouse. They can send a ping to a WMF server, but cannot be polled for data (re: Nemo, "Data sent" ;) ).
- Inhouse vs. public might be a piece of information we want to poll (could be done automatically via IP (?) or be checked). I think this information is very valuable.
- I know (at least one) very security sensitive administrator. He might be willing to send anonymous stats, but wouldn't want them to be publicly available. So I'd prefer a two-step opt in: send data and allow the data to be public.
- As Mark scetched out, this is tightly tied to the question of wiki spam.
- How about continuous statistics, like number of pages, users, edits?
I second this. I administrate a very large corporate wiki (for internal documentation) and I am sure the privacy issue will come up. I'd love to be able to configure this information in various ways (what is sent, how, etc.)
--Daniel Renfro 16:59, 30 May 2014 (UTC)
Two comments from Dan Garry Edit
This looks to be a nice way to provide some statistics about who's using MediaWiki, and where it's being used. I have two concerns, though.
- There are potential privacy implications to this depending on the nature of the information that is sent. The current resolution, namely making the person agree to send the information, does not absolve us of the responsibility of handling that data properly. Third parties are sending us data, and we should treat that seriously. In the present implementation it looks like this is handled though, as there is not really any private information sent. A declaration of what we'll use the data for would be nice.
- "Send data on your wiki to Wikimedia?" has the potential to be taken the wrong way, like we're some central authority on wikis or something. I suspect that this problem can mostly be solved by the wording of the text presented to the user, and I can potentially some support for refining the text in my role as product manager for platform. Please feel free get in touch if you'd like my help.
Comments from user Esp261 Edit
I agree with the use for easy to find/use registration for MW installs, especially in regards to keeping site operators up to date with relevant security releases and anti-spam tools they could be using. However, one potential issue with the flow of registration described: Often a new wiki will, in my experience (started one small and one mid sized wiki, helped on a handful of others including WikiApiary where I'm a sysop), want to remain non-public for a few days/weeks during the initial setup phase (adding extensions, creating basic templates, making some seed content with already trusted community members), but would benefit from registration. Having a way to either give delayed registration (i.e. "Register in one week, Register in two weeks, etc") along with clarification of possible reasons to delay (something like "register this site when you expect to have anti-spam measures configured and want it to be findable by the public") or give the option to register immediately but have the wiki non-public (does not appear on any public facing lists/searches) until an administrator alters the setting may avoid many sites slipping through the cracks due to the person setting up the install wanting to keep the site hidden initially and just not bothering to/forgetting to register it later. An alternative solution would be to move registration from install to a special page accessible only by bureaucrats and add a "register this wiki (founder only)" or similar link to the default main page. An added bonus of moving the registration to later is that the tagline and logo are more likely to be stable later on than immediately after install (even if tagline and logo are requested at the start of install, on many sites they will be revised during the "private beta" setup stage).--Esp261 (talk) 15:06, 25 October 2013 (UTC)
Some comments Edit
1) I don't like the idea of all this data being sent at installation. Installation is not a final indicator of the wiki's configuration. Lots of configuration is done after the installation process. So all that data sent – which is rather redundant given it's available from siteinfo anyways – on installation could be completely wrong. I would prefer to see minimal information sent in the ping (just stuff like api url) and then have the ping server fetch the information itself from the wiki's API after a delay. And the information being introduced here that isn't currently part of siteinfo should be added to siteinfo instead of depending on the ping for it.
2) I also like the bit about the Tagline. From what I understand MediaWiki:Tagline is not a tagline like the ones used in WordPress and other websites but a string intended to identify the wiki when the page is printed. Which is starkly different than what the attempt here is to treat the tagline as and "raise the visibility" of.
3) Also when implementing the backend I'd like to see it served from a *.mediawiki.org domain name.
- 1 — agreed, there should be opportunities to “update” the data if nothing else. Perhaps there could be a way to say “oh, you have this information — tagline? — that you didn't have before. Send it now?”
- 2 — The thanks here go to Jamie of Wikiapiary since he proposed the tagline and said it would be useful.
- 3 — The core of the backend would be hosted on a WMF-affiliated, probably *.mw.o domain name.
- — ☠MarkAHershberger☢(talk)☣ 02:52, 24 April 2014 (UTC)
Re: 1, the wiki should just send pings periodically. Configurations are updated, domain names change, corporate wikis tend to be behind firewalls; a central ping server polling for changes in every wiki is not going to work, and trying to figure out when the configuration and/or usage changed enough to send a new ping is a lot more effort and a lot less robust than just sending it every week. --Tgr (WMF) (talk) 01:37, 3 December 2014 (UTC)
It might work better if you offered something in return Edit
I'm pro Opt In registration and/or tracking. It helps a lot to know how many people are using (say) PHP 5.2 and so on. WordPress does it via their API in order to perform plugin updates, which is a feature I've always wished MediaWiki had. Upgrading extensions has always been annoying unless you use git, and even though there's no way to 'know' if you need an update. With that in mind, it would likely be more appealing to the common user if there was some benefit besides 'You're tracking me to improve your stats and make things better.' While that logic is perfectly understandable to devs, I see a lot of people in the WordPress plugin world ask for opt-in tracking and the majority who get it are giving the users a clear benefit back. As I mentioned, WP does this to provide updates. Okay, if you could use the data to tell people on the Special:Version page that an extension was out of date, that would be an immediate benefit and it could help security. Otherwise, the majority of your non-technical users (that is the people you want to track the most) won't bother.
Obsolete RfC? Edit
This was already done (although not as extensively) when Manual:$wgPingback was added.