Page MenuHomePhabricator

SiteMatrix cleanup
Closed, ResolvedPublic

Description

The following codes appears at Special:SiteMatrix yet no project exists with that code or has been deleted:

  • cz
  • dk
  • epo
  • jp
  • minnan
  • nan
  • nb
  • zh-cfr

The following codes appears at Special:SiteMatrix but do not have a language description attached to it:

  • bug

Is a cleanup of that page possible? Thank you.


Version: unspecified
Severity: normal
URL: https://meta.wikimedia.org/wiki/Special:SiteMatrix

Details

Reference
bz27194

Event Timeline

bzimport raised the priority of this task from to Low.Nov 21 2014, 11:15 PM
bzimport added a project: SiteMatrix.
bzimport set Reference to bz27194.
bzimport added a subscriber: Unknown Object (MLST).

(In reply to comment #0)

The following codes appears at Special:SiteMatrix yet no project exists with
that code or has been deleted:

  • cz
  • dk
  • epo
  • jp
  • minnan
  • nan
  • nb
  • zh-cfr

The following codes appears at Special:SiteMatrix but do not have a language
description attached to it:

  • bug

Is a cleanup of that page possible? Thank you.

This seems to be a problem with the secure server. On the regular servers, those "nonexistent" language wikis redirect to other wikis; that is, cz is an alias for cs and dk is an alias for da.

I believe a good solution to this issue is to remove these codes from [[Special:SiteMatrix]] totally (I wonder why they are there in the first place).

This appears to be related to bug 19986, since most of the redirected domains were set up temporarily for the purposes of moving the databases to the correct language code. If that bug is resolved, perhaps it will clear some of the confusion associated with the "nonexistent" langcodes. But unfortunately, it will take some time to get through.

For anyone wanting to do this, http://noc.wikimedia.org/conf/langlist can be copied into a local file for the SiteMatrix site list file and $wgSiteMatrixFile can point to it. This allows for some testing. This is easy to set up.

Related URL: https://gerrit.wikimedia.org/r/60329 (Gerrit Change I57d4a6d794cbb1cc746a08d3a4289431c768b5a1)

Seem the API handles this a bit more cleanly: https://en.wikipedia.org/w/api.php?action=sitematrix

<language code="cz" name="">
  <site />
</language>

We can't just remove them from langlist as the file is used by the DNS configuration.

cz, dk, epo, jp, minnan, zh-cfr don't have language names defined.

nan and nb do have language names defined

Is it worth just having a blacklist for Special:SiteMatrix? (see above)

(In reply to Sam Reed (reedy) from comment #6)

We can't just remove them from langlist as the file is used by the DNS
configuration.

Does it require input from ops then?

Is it worth just having a blacklist for Special:SiteMatrix? (see above)

What blacklist would you have in mind?

Glaisher claimed this task.
Glaisher added subscribers: Reedy, Glaisher.

We can't just remove them from langlist as the file is used by the DNS configuration.

A separate file is now used for DNS. There is also an open patch at Gerrit for removing these codes from the langlist file.

I'll close this task as the patch above by Reedy (which was merged on 2013) skips these codes and there is a local name for bug as well.

Glaisher set Security to None.
Glaisher removed a subscriber: Unknown Object (MLST).

https://gerrit.wikimedia.org/r/#/c/166281/ has been merged today so the underlying issue has also been fixed now. :)

Shouldn't that blacklist be configurable?