On Sat, 31 Aug 2024 15:05:58 +0200 Ekaitz Zarraga <eka...@elenq.tech> wrote:
> Hi, > > On 2024-08-31 14:05, Jan Wielkiewicz wrote: > > Hello again, > > > > On Sat, 31 Aug 2024 01:45:04 +0200 > > Ekaitz Zarraga <eka...@elenq.tech> wrote: > > > >> Hi Jan, > >> > >> > >> First, there are tags in the issue system: > >> > >> https://issues.guix.gnu.org/search?query=tag%3Aeasy > >> > > > > Missed this completely. Where is this documented? > > In the bottom of the search box, you can show hints and then it > suggests to use them. > > >> > >> The issue of friendliness has been brought up in the past, I > >> suggest you to search in the mailing list for previous > >> interactions in the topic. > >> > >> In fact, I took part on a discussion about more general topics and > >> I mentioned we should be more empathetic with other people's needs: > >> https://lists.gnu.org/archive/html/guix-devel/2023-09/msg00488.html > >> > > > > Okay, will read this later. > > > >> > >> Tedious and counterintuitive are very loaded words, and not very > >> objective either. The intuitiveness of things is based on the > >> previous context people has. This is a GNU project, and the context > >> for many GNU developers is autotools, mail based workflow and so > >> on: maybe for GNU developers that's the most intuitive thing. Who > >> should we please? > >> > > > > I used these words because they convey my feelings related to the > > current process. The discoverability of the system is low - every > > time I send patches I need to reread documentation to then finding > > out my email provider does not cooperate (failing SMTP, having to > > re-enable email client support to proceed). So I probably need to > > change my email provider in order to contribute to Guix smoothly. > > This was literally my situation not that long ago. I'll explain more > below. > > > I'm aware this is a GNU project, I'm a free software enthusiast and > > GNU/Linux user myself, but I'm not an epic GNU hacker and you can't > > expect all contributors (who simply want to update a package they > > use) to be epic GNU hackers. > > I agree with this. We are not expecting that either. That's why I was > trying to push for a better documentation. I read your original > message as coming from frustration and I don't think you were very > empathetic when you wrote it. > > >> In my case, I find the Github-like PR based approach quite > >> counterintuitive, but I got used to it before collaborating in Guix > >> and now I don't dislike it. Starting using PRs for everything made > >> me a little bit reluctant to try the email based approach. > >> Surprisingly, it happens to be way easier than it looks, and it's > >> actually quite cool. > >> > > > > Wasn't easy to learn it at first, but running "git push origin" to > > update a remote branch linked to a PR is pretty straight-forward. > > As simple as `git send-email HEAD~`. Our way is not more complex, > it's just different. > > >> You should give it a go: https://git-send-email.io/ > >> (If you don't like or you can't use that, you can paste the patches > >> in an email, or send them as attachments, which is as difficult as > >> taking part in this mailing list.) > > > > I have this installed and use it, can't send patches to Guix > > otherwise... > > The link is a very good tutorial on how to use git-send-email. I like > to share it because it made a huge difference to me. It's just a very > short tutorial that explains very well the workflow in a practical > way. > > >> Once you use it for as long as you used other options, I'd love to > >> have this conversation with you again (I'm not saying you'd change > >> your mind, but the conversation would be fairer). > >> > >> I will share some thought here that I know many people in the > >> community doesn't agree with (and I'm kind of a 50-50 on it) but I > >> think it's worth sharing: > >> > >> > >> So, it is true that making the access to things harder than the > >> possible minimum is simply stupid. There's no discussion there. > >> That being said, it would be very sad if a seasoned software > >> developer, who proved to be able to write code for a complex > >> project like Guix, is not able, is discouraged to, or is afraid of > >> sending the patches via email (but is ok opening an account, > >> making a fork, making a pr and so on). > > > > I'm apparently smart enough to install Guix as my daily driver > > distro, learn Emacs with Paredit and hack together packages, and > > then an email-based system filters me out... > > This is a very interesting point and I encourage you to dive deeper > on it. In my case I had pushed for a less emacs-centric approach in > the docs, and I wrote some parts of the documentation for people that > used other editors. I'm not smart enough to use emacs, and I gave up > on it, but with the tutorial above I overcame my fears with the email > based approach. > > >> Putting it in a different way: not every contribution is worth > >> attention. If people who want to contribute are not able to send an > >> email, I don't think their potential contributions would be that > >> interesting or worthy. The skills needed to be able to participate > >> in Guix are quite advanced, I don't think the email is a burden for > >> people that have them. > >> So let's say, yes: we do migrate to codeberg.org. And now what? Do > >> you think the amount of contributions will skyrocket? Do you think > >> the amount of meaningful contributions will be higher? > > > > Guix is not one thing - updating packages is not the same kind of > > task as for example hacking on low-level system or package manager > > functionality. I believe simplifying or making contributions more > > similar to what other projects have would indeed help. One example > > would be a free software game project I take part in. The original > > founder of the game did not know git and maintained the game using > > GitHub's online editor. Then when he stepped down, we moved the > > project to Codeberg. In the very same project we also taught a > > literal Windows user how to use git and command line which then > > enabled him to make cool and meaningful contributions. And then I > > myself did several contributions to Guix while barely knowing what > > I'm doing. So I'd be careful with underestimating the power of > > inexperienced users. > > Yeah, I can agree with you, but we should as a whole make a > cost-benefit analysis of this. Maybe making our current audience > uncomfortable just for potential contributions that we don't know how > many they can be is not a good idea. We should think about this > deeply, and I opposed your email just because it was too categorical > and I wanted to share the other side of the coin with you. > > >> I know what I say here might smell like some intellectual classism > >> or something, but if we slow down a little bit: implying that > >> there's a lot of people wanting to contribute to Guix and they > >> cannot overcome the burden of the tooling we use is a little bit > >> condescending to them. > > > > I'm doing my best to overcome the burden of tooling, but in the very > > same time I could just make my own channel with software I want > > updated on Codeberg, instead of contributing them to upstream Guix. > > > > This is fine, Guix's nature does not only allow this but also > encourage it. In the moment you want to upstream your changes, you'll > have my help in the IRC or in the Mailing List to guide you over the > process. > > >> > >> *People are smart*. They'll figure out. Our process could be > >> better, specially the documentation for newcomers (which btw > >> improved a lot recently, and mumi is great), for sure, but, in > >> summary: > > > > Yeah, the current format of the document makes it easy to overlook > > important things, for example there's no header called "Canceling > > Your Issue", "Updating Your Patch Series", etc. Codeberg's user > > interface makes these things discoverable under nice green/red > > buttons. > > > > This is way more constructive than your previous email and is very > interesting. Suggestions like this help a lot to make a better > documentation and become more welcoming and are not radical changes > that require a huge agreement in the community. This we can do, right > now. > > About those specific issues, debbugs allows to close the issues by > email. Guix documentation shows how to do it with mumi: > > mumi compose --close > > But I agree we could do better, and I suggest you to point out all > the things you'd like to have and I commit myself to updating the > documentation with them. > > >> > >> - I don't dislike the PR style > >> - I doubt if using a PR based approach would result in more and > >> better contributions > >> - Just because something is more popular that the other thing it > >> doesn't mean the latter is less intuitive > >> - Our process is not more complex, but it might be *new* for people > >> - If we provide GOOD documentation, people will understand it > >> - We have communication channels (that work) for the cases when > >> people gets stuck > >> > >> > >> Those are my thoughts. > > > > In the end you could argue that it's my own fault for using a > > broken/incompetent email provider, but then it's the contribution > > system of Guix that exposed this issue as opposed to alternative > > systems like web GUIs that usually just work. Perhaps adding "your > > email provider may be broken, try using a different one" and "make > > sure to enable support for a standalone email clients in your email > > settings" to the documentation would help? I spent more time than > > needed in my life looking for solutions to fix email sending. > > > > I have had SMTP problems with my previous email provider (protonmail) > and I had to paste the patches in the body of the emails, or add them > as attachments. It worked, but this was one of the main things that > prevented me from trying git-send-email. > > I am familiar with your issue, and it's really disappointing. I > finally changed my provider and my quality of life was improved > dramatically. > > If you don't want to change your provider, which is perfectly ok, you > are probably going to find some things a little bit more > uncomfortable, but it's still possible. > > When I had issues with all this, people in the Guix community helped > me with the setup, corrected my patches, and encouraged me to be do > it better next time. I wasn't scared off by people: nobody here wants > that. What I got were some mentors that encouraged me to give the > best of myself, not only in this project but also in others. > > Please correct me if I'm wrong but from your previous message I read > you were frustrated and maybe angry about it. From this message I > read what you need is just some support. I think I can help. If you > want, we can guide you to be more comfortable with the platform and > document everything that made your life easier in order to make next > person's approach to Guix smoother. > > > So yeah, that's pretty much it > > What do you think? > > Should we start? > > I apologize for my emotional and not constructive reply at first, I'll try making more constructive replies next time. As for help, now I'm good because I managed to send patches correctly-ish. I'll probably end up looking for another email provider. The issue that needs to be addressed however is the documentation, this page to be precise: https://guix.gnu.org/en/manual/devel/en/html_node/Sending-a-Patch-Series.html I propose turning it into numbered list of things a contributors needs to do in order to successfully sending 1) a single patch or 2) a patch series, with clearly visible headers. Then as I mentioned in the previous reply, sections such as "Canceling Issue", "Adding More Patches", "Updating/Changing the Patch Series" and "Troubleshooting" should be added. This should make the guide easier to follow and more friendly. The "Troubleshooting" section should cover typical mistakes/errors and mention the issue we were having. That way possible contributors will at least have a hint what could be wrong with their setup, which should limit the level of frustration. You could also add "make sure to ask on the IRC or mailing list first if you can't resolve the issue". That should do. -- Jan Wielkiewicz
