I don't think we should make new links to git.wikimedia.org, and I'm not sure a link to those gerrit project admin pages would be too useful.

I'll add that linking back to our repositories would also help getting the canonical location visible in search results, while currently GitHub can get the canonical location hidden.

https://support.google.com/webmasters/answer/66359 said (bold mine):

Syndicate carefully: If you syndicate your content on other sites, Google will always show the version we think is most appropriate for users in each given search, which may or may not be the version you'd prefer. However, it is helpful to ensure that each site on which your content is syndicated includes a link back to your original article.

Yes please. The current descriptions are halfway between useless and actively hostile.

What format would you need? A script for generating the text or something that actually interacts with GitHub?

In T109939#3063116, @Tgr wrote:

What format would you need? A script for generating the text or something that actually interacts with GitHub?

Both, but mostly the latter, like I said I'm not going to run around editing them by hand :)

I just assumed something existed already (how did these descriptions get created in the first place?)

I used to have a plugin that created those repos automatically (with a description), but it stopped working and I lost the source code. Now repos are created manually on github.

In T109939#3063152, @demon wrote:

I used to have a plugin that created those repos automatically (with a description), but it stopped working and I lost the source code. Now repos are created manually on github.

Thanks for the context. Some pointers for myself:

• https://developer.github.com/v3/repos/#edit
• https://github.com/PyGithub/PyGithub

git.wikimedia.org has perished, and the gerrit admin page and the Phabricator repo page both utterly suck as a landing page. I think it would be better to use the description as it's normally intended, use the URL to link to the project page on mediawiki.org (or leave it empty if there isn't one), and put everything else in CONTRIBUTORS (see T136863#3714110).

Other aspect: Having no way to differentiate repositories mirrored from Gerrit vs. repositories whose canonical home is on GitHub makes it impossible to get proper developer statistics for Wikimedia projects canonically hosted on GitHub.

In T136863#3947933, @demon wrote:

I actually think we shouldn't bother replicating everything....there's a ton of crap.

What I want to do is re-write my old github plugin again (the upstream one I don't like). It'd handle repo creation (as appropriate), along with keeping descriptions in sync.

We could use github repo topics to identify the repos which are active on github.

In T109939#3724744, @Tgr wrote:

[...] the gerrit admin page and the Phabricator repo page both utterly suck as a landing page.

This seems like a blocker, none of the alternatives are good:

• https://gerrit.wikimedia.org/r/eventlogging - the raw repo, but cannot be browsed.
• https://gerrit.wikimedia.org/r/#/admin/projects/eventlogging - obscure admin page, but with fairly obvious cloning instructions
• https://gerrit.wikimedia.org/r/plugins/gitiles/eventlogging/+/master - the most GitHub-like UI
• https://phabricator.wikimedia.org/diffusion/GEVL/ - another nice UI, but itself a mirror

I think it would be better to use the description as it's normally intended, use the URL to link to the project page on mediawiki.org (or leave it empty if there isn't one), and put everything else in CONTRIBUTORS (see T136863#3714110).

I like this suggestion, once the user navigates to a proper landing page, it's reasonable to include bare gerrit repo URLs for example.

In T109939#4217488, @jayvdb wrote:

We could use github repo topics to identify the repos which are active on github.

I think this is a good idea. A few weeks ago when I was cleaning up things in the https://github.com/toolforge/ org account I went through and added a "mirror" topic tag manually to the repos there which are mirrors of Diffusion or Gerrit upstreams.

Some bot could be written to walk the repos under the wikimedia org account and add "mirror" based on the presence of a .gitreview file in the repo (criteria used by @majavah in T249703). @majavah even volunteered at T249703#6157963 to write such a bot to take care of this task if folks could just settle on the text to be used.

In T109939#3724744, @Tgr wrote:

git.wikimedia.org has perished, and the gerrit admin page and the Phabricator repo page both utterly suck as a landing page. I think it would be better to use the description as it's normally intended, use the URL to link to the project page on mediawiki.org (or leave it empty if there isn't one), and put everything else in CONTRIBUTORS (see T136863#3714110).

• The mediawiki/core repo uses this hand-edited description: "🌻 The collaborative editing software that runs Wikipedia. This is a mirror from gerrit.wikimedia.org. See https://www.mediawiki.org/wiki/Developer_access for contributing."
• The ops/puppet repo uses this hand-edited description: "Wikimedia Foundation operates some of the largest collaborative projects in the world. This is our Puppet repo. This repository is a mirror; see https://www.mediawiki.org/wiki/Developer_access for contributing."

Both avoid the problem of figuring out what Gerrit landing page to use in the description which is inline with @Tgr's suggestion at T109939#3724744. The default for a repo could be Gerrit project's description if it has one (see https://gerrit.wikimedia.org/r/#/admin/projects/?filter=mediawiki%252F for examples) followed by either "This repository is a mirror; see https://www.mediawiki.org/wiki/Developer_access for contributing." or "This is a mirror from gerrit.wikimedia.org. See https://www.mediawiki.org/wiki/Developer_access for contributing.".

Figuring out what project page to link to might be a bit tricky, but there are certainly heuristics that could be used to guess for many of them. If the bot can't figure it out, it could just leave the url blank.

The CONTRIBUTORS aspect is reasonable as well, but I think it is trickier to deal with as it involves changing the repo contents rather than just the metadata provided on the GitHub side. I saw enough angry push back against the CODE_OF_CONDUCT.md additions to stay away from forcing content into existing repos myself, but it seems reasonable that we could at least have a wall of shame tool to show repos that are missing a CONTRIBUTORS file and a suggested default.