Proposal: migration of Parabola infrastructure over to a self-hosted instance of SourceHut

gap - 21 days ago -

SourceHut uses relatively lightweight modules over a monolithic source forge.

The applicable modules of SourceHut could replace the Redmine instance powering the forum and issue tracker, the cgit instance powering the web interface for the git repos, the Mailman instance powering the mailing lists, and even replace the wiki with its web accounts and markup syntax with git commits, mailing list pull requests, and markdown files rendered statically.

We could also utilise its extra features such as CI for improving the quality assurance of our projects.

According to the GNU ERC [1] and the GNU ERC evaluations [2], SourceHut has multiple issues, although it's nothing we can't fix in a libre build in the libre repository and by enforcing licensing and labelling policies.
Moreover, based on the history and willingness of its main dev to be evaluated on the GNU ERC [3], the SourceHut project appears to be more likely than most to accept our liberation changes upstream.

The main problem I can see is that we would need to remove support for CI for systems which are not 100% libre, which would amount to removing all of them, and then we'd have to add Parabola so as not to have a useless feature.

One potential issue for the moment is that SourceHut does not officially label itself as "stable" although it's been in development and has been used daily by hundreds of projects for years.


Replies (2)

RE: Proposal: migration of Parabola infrastructure over to a self-hosted instance of SourceHut - bill-auger - 9 days ago -

parabola users can already contribute to abslibre on the pagure forge - if a significant number of people start contributing to it, the plan was always to self-host an instance eventually

pagure was chosen specifically because it was determined, that it's maintainers were the most libre-minded, during a thorough research project we did for notabug - pagure is also more user-friendly that sourcehut - IMHO, sourcehut is trying to fill a niche which does not exist - the very people who would prefer it over others, are those who would prefer not to use any web forge, but offer one for people who do - unfortunately, people who prefer web forges, would be very dissatisfied with sourcehut, for the very same reasons that others may prefer it (its simplicity, lack of javascript, , shiny "click-me" buttons, and other fancy "modern" features such as "push-notfications", "@mentions", "likes", etc)

personally, i am in the former group; and i believe that most parabola devs are - i appreciate that sourcehut is super-fast, light-weight, and the interface is plain (not "noisy") - but the only reason for parabola to have a forge, is for people who prefer fancy webby tools; so my preference is not relevant to the decision

we did a lot of research on this, with input from people of all experience levels; and people preferred gogs/gittea and pagure, by a significant margin, over the dozen or so others which we evaluated

to be clear, the only projects which actually need a forge, are those who can not, or will not host their own infrastructure - for example, around the same time we setup the pagure repo, debian started hosting a forge for people who prefer using them - but it was not intended to be the primary tool for debian development, and is not used for bug reports - debian still uses email, FTP, and tarballs primarily, and probably always will

this would not be a case of "migration of Parabola infrastructure" - it is more of a "supplement", to make contributions easier for people who are daunted by email

RE: Proposal: migration of Parabola infrastructure over to a self-hosted instance of SourceHut - gap - 8 days ago -

The way I see it, SH is the best of both worlds: the hackers will be happy there is a minimalistic web interface which doesn't need JS and integrates very well with their existing mail-driven workflow, and the non-technical people will be happy that they have a web interface to scroll through and click about on without using email, although ultimately all real work is transmitted via email.

The heavyweight Github-inspired forges all heavily rely on web interfaces, JS, and resource-intensive server-side applications, which is a slower and clunkier workflow experience, and is totally irrelevant if the real work is going to be ultimately transmitted via email anyway.
At the end of the day, we could always have a read-only Gitea mirror of the SH repos for those who prefer that particular interface.

I think the only reason those particular interfaces are presently preferred is because people are already used to them due to Github, and it wouldn't be a particularly difficult learning experience to switch to SH instead.
Moreover, whilst it's already stable enough to adopt today, SH is still technically in an "alpha" stage of development, which I suppose is a good excuse to say that the UI might still change before it becomes officially deemed "stable".