Disable a project without deleting it?

[shoyer]

@rgommers and I are maintainers for NumPy.

There is currently a broken RTD config for NumPy that is attempting (and failing) to build every new commit. This seems like a needless waste of resources -- and also users could stumble upon the broken docs -- but there is concern (numpy/numpy.org#431) that if we deleted the project someone unaffiliated could register the name.

Is there some way we could disable the NumPy project without deleting it, or otherwise achieve this goal?

Thanks,
Stephan

[astrojuanlu]

We do have a policy for abandoned projects modeled after PEP 541, but we don't seem to have a policy on name squatting. I also looked into using redirections, but I don't think it's possible to redirect to an internal external website. Thoughts @readthedocs/core ?

[stsewd]

You can create redirects to external sites, but redirects only work on 404s. So in order to have a global redirect you can import an "empty" project (with only an index.rst), set an exact redirect from /en/latest/$rest to https://numpy.org/doc/stable/ that would redirect any 404 like numpy.rtd.io/en/latest/foo.html to https://numpy.org/doc/stable/foo.html, additionally you would also need to set a redirect in your index.rst manually with some html or js <meta http-equiv="Refresh" content="0; url='https://numpy.org/doc/stable/" />.

Note that you can omit $rest from the redirect if your new site doesn't follow the same estructure.

docs on custom redirects https://docs.readthedocs.io/en/stable/user-defined-redirects.html#exact-redirects

[shoyer]

I also tried to disable doc builds for "numpy" as well, because as far as I can tell RTD is still trying (and failing) to build the docs upon every commit.

A re-direct isn't really so important here -- a 404 would be fine.

I guess the simplest solution would be to delete the project, and just keep an eye out for anyone claiming the name. If that comes up, I assume the RTD team will be responsive.

[stsewd]

I also tried to disable doc builds for "numpy" as well, because as far as I can tell RTD is still trying (and failing) to build the docs upon every commit.

You should remove the webhooks that are pointing to rtd from the github project.

[shoyer]

Oh nevermind -- someone just manually tried doing a build a few hours ago. There are no web-hooks to worry about :)

…

On Wed, Apr 28, 2021 at 7:04 PM Santos Gallegos ***@***.***> wrote: I also tried to disable doc builds for "numpy" as well, because as far as I can tell RTD is still trying (and failing) to build the docs upon every commit. You should remove the webhooks that are pointing to rtd from the github project. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#8143 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAJJFVXEPTRCCKHSO5SY2LTTLC5C5ANCNFSM43X5JWWA> .

[humitos]

Some ideas:

• remove all the webhooks from GH
• remove the webhooks integration from RTD
• disable all the version on RTD
• change the repository on RTD to a template one
• import a empty project with sphinx-notfound-page installed (https://sphinx-notfound-page.readthedocs.io/)

I'd personally add a "nice index" saying that the project was moved but it's kept to avoid "stealing the namespace" and confusing users. Why I am saying this? Because we do have a policy for abandoned projects and if someone request numpy namespace and I'd see a 404 project in a year or 2 I'd mark it as abandoned and free that name.

[humitos]

The bigger questions are:

• why is it failing on Read the Docs?
• does it worth solving that issue and use Read the Docs as the canonical source for the docs?

I suppose that you already talked about this, tho.

[rgommers]

disable all the version on RTD

Looks like you cannot do that, there's Active/Hidden buttons and for latest you cannot disable Active if nothing else is left.

why is it failing on Read the Docs?

that's not relevant, we do not want to build or host on RTD - we just also do not want anyone else to take over the numpy name.

[rgommers]

I'd personally add a "nice index" saying that the project was moved but it's kept to avoid "stealing the namespace" and confusing users. Why I am saying this? Because we do have a policy for abandoned projects and if someone request numpy namespace and I'd see a 404 project in a year or 2 I'd mark it as abandoned and free that name.

This looks like a good idea.

[shoyer]

I'd personally add a "nice index" saying that the project was moved but it's kept to avoid "stealing the namespace" and confusing users. Why I am saying this? Because we do have a policy for abandoned projects and if someone request numpy namespace and I'd see a 404 project in a year or 2 I'd mark it as abandoned and free that name.

IMO it would be worth reconsidering for cases where the name is already "claimed" in the Python community by another project and could potentially be a source of confusion. Kind of like how Pypi doesn't let you register projects with the same name as standard library modules.

[shoyer]

why is it failing on Read the Docs?

that's not relevant, we do not want to build or host on RTD - we just also do not want anyone else to take over the numpy name.

For context, it was challenging to get NumPy's doc generating working on RTD for a few reasons:

1. NumPy is mostly C code, and needs to be compiled from scratch (in each build) to compile properly.
2. The docs also need other messy to build libraries to be installed (e.g., matplotlib), which in turn depend on NumPy!
3. The doc building process is also slow.

Presumably these could all be fixed, but I'm not sure it would be a good use of anyone's time. We already have a pretty good custom CI setup for building NumPy's docs, and NumPy is a bit of a special case (as the foundation of the numerical Python ecosystem).

[astrojuanlu]

Thanks a lot for the extra context @shoyer!

[ericholscher]

The main challenge here is that RTD is larger than Python. So there's a large number of projects in other package indexes which have similar names to Python ones, and we can't really reserve every special case in every language.

I think the proper solution is just changing the numpy project on RTD to an invalid git URL and it will never build. The better long-term solution is for us to have a repo like readthedocs/reserved that you can put into the git URL, and it generates a nice index page. This wouldn't be hard to do. We could then have a small guide on it.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[humitos]

@ericholscher I saw that you marked this issue as Accepted. Is there anything actionable from our side to do here?

[ericholscher]

I think probably a guide: Migrating from RTD to another hosting provider or something like that. Not a huge priority of course, and there might already be an issue open for it.

[astrojuanlu]

I'm confused, isn't this about being able to "reserve" or "disable" a project?

[stsewd]

@astrojuanlu I think both apply :D like you can create an empty project that redirects to the new domain, that means leaving the old project always reserved/active.

[humitos]

We have implemented "Force redirects" now and we also wrote a small guide about how to migrate your docs to a different domain: https://docs.readthedocs.io/en/stable/user-defined-redirects.html#migrating-your-documentation-to-another-domain

I'm closing this issue, but feel free to re-open if there is another action that Read the Docs should do at this point. Also, let us know if you find that guide useful to achieve your goal 👍🏼