[Standards] Evaluating gitlab.com as new location for XEP Editor repositories (xeps+registar)

Sam Whited sam at samwhited.com
Sun Jun 14 15:17:44 UTC 2020


Ahh, I see, this is a Docker Hub issue. I completely agree that Docker
Hub has not been a success. It is also quite easy to build Docker
images with other CI platforms though while remaining on GitHub (ie. I
think you already submitted a PR to use a GitLab CI build). I've built
Docker images with GitHub actions (my preferred solution), Sourcehut
builds (would be my preferred solution, except that it's still very
*very* alpha and I'd hate for people to have yet another account just
for CI), and GitLab CI in fact (which I was not a fan of but I'm not
the one configuring it or having to deal with it for now so that
doesn't really matter).

Can we not build Docker images with one of those platforms while leaving
the actual repo where it is? If so, that seems to be the direction we
should go in first. If the board then wants to decide we need to move
all the repos again for some other reason, that's fine but this doesn't
seem to call for that.

—Sam

On Sun, Jun 14, 2020, at 11:10, Jonas Schäfer wrote:
> Hi Sam,
>
> Thank you for your feedback. I hear you. I totally see where you’re
> coming from. Moving off GitHub, which is, as you say, where the people
> are, is a thing which needs to be considered carefully. Ultimately,
> the Board will have to decide (hence also CC to members@).
>
> First of all, please let me say that I brought this up purely for
> technical reasons. I don’t have any horses in this race on whether
> we’re on GitLab or GitHub, for political or other reasons (I don’t
> believe much in the GitHub-is- now-microsoft-and-should-be-avoided
> thing [1] and I think we’re not quite there yet with decentralised or
> de-monopolised code hosting. So this is truly not it. Though that hype
> happens to have given us control over gitlab.com/xsf, so there’s that
> at least.).
>
> I’d prefer to let things be with the status quo, even if only because
> updating all the references will be a pain, if the status quo wasn’t
> terrible.
>
> Originally, this wasn’t about xeps even. This was about the registrar
> repository, which has been malfunctional since the last major change
> to the infrastructure back in 2017(?). That is an unacceptable state.
>
> We’ve been collaborating with iteam (which also is short on time,
> mind) to find a solution. As it turns out, the current interaction
> between Docker Hub and GitHub cannot be replicated without introducing
> security problems (giving Docker Hub write access to all xsf
> repositories) due to changes on either side of the APIs involved.
>
> In addition, we need a solution which is as simple as possible on our
> end of things, which means ideally that it works with just a docker
> pull && docker run.
>
> Thus, I’ve pondered other options. GitLab CI has the massive advantage
> of being tightly integrated with the GitLab platform. I was able to
> set up the registry and xep builds within just an hour or two and have
> been improving on them since then, giving us automated emailing and
> archiving.
>
> It also seems to be faster on average than the average Docker Hub
> build, which is good for editor tiredness. Please see the original
> email for notes on which immediate advantages we gain from what I just
> did this weekend on the editor side of things.
>
>
> > Also just in general, I completely disagree. We need to be where the
> > people are, and the people are on GitHub whether we like it or not.
> > Don't split up the repos and make XSF resources even harder to find
> > than they already are, don't make things more complicated than they
> > need to be, and everything will be fine.
>
> The exodus of editors since the last change to the infrastructure
> (moving the builds to Docker Hub) makes me think that it is very much
> not going to be fine. pep. and I are the two people doing most of the
> day-to-day work in the team (out of the six members we still have
> left) and it’s still draining. Calls for support have gone (mostly)
> unanswered.
>
> This is fixing the issues in the Editor work you complained about in
> the past and which, I think, caused you to leave the team.
>
>
> > Please don't do this.
>
> I don’t see how we have a choice unless someone finds a way to do all
> of what I mentioned on GitHub. Or at least the most important parts of
> that, which would be the automated tagging, archiving and improved
> build times, and all of that without introducing a security nightmare.
>
> The Registrar has been unable to operate for over three years now.
> Changes have been blocked and rejected because of this. It is not a
> way we can continue to operate.
>
> Sure it would be nice if we could stay at a place where "the people
> are" (though I think the ability to log in via GitHub makes the hurdle
> rather small for GitLab.com). That is indeed the thing which makes me
> most uneasy with this change. However, I’m more uneasy with not having
> working Editor processes for such a long time and seeing the standards
> work suffer under it.
>
>
> kind regards, Jonas
>
>
>  [1]: https://sotecware.net/on-centralisation-of-code-hosting-infrastructure-an-argument.html
> _______________________________________________
> Standards mailing list Info:
> https://mail.jabber.org/mailman/listinfo/standards Unsubscribe: Standards-
> unsubscribe at xmpp.org
> _______________________________________________
>
> Attachments:
> * signature.asc

-- 
Sam Whited


More information about the Standards mailing list