A (Partial) Defense of Debian
changelog.complete.org
changelog.complete.org
The section on personal preferences is one example of this. Very few people new to the project would prefer debbugs and a text-based mail client over a Github workflow.
edit: I may have read that part of the post uncharitably, if what the author meant was continually switching between the browser and the command line. Still, the sentence stuck out to me.
Realistically, I don't see them switching away from text based files with at least the option of an offline workflow on the back-end, which makes sense to me. That said, I really do wish they'd understand that to people who don't live and breath the Debian workflow, the existing front-end tooling is painful to use and antiquated and prioritize updating it. It is possible to create a much nicer front-end user experience on top of what currently exists.
[1] From the FSF/Debian perspective, no-cost is not free https://en.wikipedia.org/wiki/Gratis_versus_Libre
Pagure provides a service similar to GitHub and GitLab, while also supporting mostly offline workflows. Pull request metadata, issues, documentation, and of course the code, are stored as Git repos. You can clone them, do work, and push them back, and Pagure will act on them appropriately. Or you can operate from the web UI if you prefer, and ignore the fact that the data is stored as Git repositories internally.
HyperKitty adds web forum-style conversation views and functionality to a mailing list, allowing people who are uncomfortable with handling everything by mail to still follow and respond to threads and be part of the conversation.
Both of these show it's possible to support the old guard while making it possible to welcome fresh blood and adapt to their preferences.
For better or worse, it's an "adapt or die" situation. Debian isn't adapting, which is not good.
As an aside, Debian nearly did deploy Pagure. The community voted for it, and then the DSAs (Debian System Administrators) decided on using GitLab using GitLab Omnibus (not even the GitLab packaged in Debian!) anyway, which upset a fair number of friends I have who are Debian Developers. But all of them said that they were powerless to influence change, which seriously speaks to some kind of broken process here, where the "most powerful" group of people in Debian feel powerless about stuff like this.
It's also annoying to have to be on the internet just to do development work, but that's a lesser complaint.
Maybe it will in the future, but it doesn't now. New members added in 12 months (data drawn from https://nm.debian.org/public/people):
2018-03..2019-02 80 new members
2017-03..2018-02 59 new members
2016-03..2017-02 87 new members
2015-03..2016-02 68 new members
2014-03..2015-02 49 new members
You control the operating system. Couldn't you build a web form that integrates with a local agent?
In addition, we created the Automatic Bug Reporting Tool (ABRT)[1] to handle program and system crashes and report them so developers can look at them.
We have hooks in various parts of the system to help generate useful bug reports of serious issues with crash data, machine information, and so on. They get submitted as private bugs by default to the bug tracker, and the Fedora retrace server[2] kicks in to request the debuginfo data remotely to get traces as needed and include those in the bug report. ABRT also helps you strip sensitive data (if any) from information being sent remotely before sending it.
You can see pretty charts about the crash analysis too: https://retrace.fedoraproject.org/faf/summary/
> Go to the website for the thing, click fork
> Now clone that fork or add it to my .git/config, hack, and commit
> Push the commit, go back to the website, and submit a PR
> Github’s email integration is so poor that I basically have to go back to the website for most parts of the conversation. I can do little from the comfort of mu4e.
> Remember to clean up my fork after the patch is accepted or rejected.
> Compare that to how I’d contribute with Debian:
> Hack (and commit if I feel like it)
> Type “reportbug foo”, attach my patch
> Followup conversation happens directly in email where it’s convenient to reply
I've actually seen this type of thing quite often in criticisms of GH/web-based workflows, etc. And frankly, they're all complete madness. The leaps of logic made in the above bullets are so enormous they can't not be wilful. The author is really stretching.
1. The equivalents of "going to the website", forking, adding to local git config, are all skipped for the Debian workflow; it is assumed one has already set the project up locally. For a new project, these steps are 100% guaranteed to be more complex for Debian (or any project not on a website with fork-like features).
2. "Github's email integration is so poor": why have we brought up email here. Github integrates PR links into the server response of the git push itself. You have actual cli integration here, shown in the output of git. Why bring up email?
3. Pointing out that you should clean up branches in your fork is superfluous.
4. The second step in the Debian workflow leaves out two things: you haven't included the setup of the reportbug cli, and—since we're clearly more comfortable with a cli than with a browser—ignores the fact that there are equivalents for interacting with the GH API without ever going near a browser.
That feels like an unfair remark, creating a pull request is only the first step in either workflow, having good email integration is still important for every other step
Cleanup for forks isn't always great, I can admit that, but you can create an org for temporary stuff, and just nuke everything there on occassion and put your forks there for temps.
Naah, you only need to know the URL of the Debian package's .dsc file that you want to patch, and pass it to the dget command, it will download the application tarball, the Debian patches and the dsc file.
You can find the dsc URLs on the Debian packages tracker site: tracker.debian.org
On a more serious note: you open with "Naah", but then you do go on to describe steps that the author neglected to include in the bullet points. My only point was: the author goes out of their way to make the comparison imbalanced.
In reality it's even simpler than what I described (if you don't remove the deb-src lines from your apt source config like I do). If you want to modify a Debian package you only need to use "source" instead of "install" for the package you need, for example "apt source curl" will download the tarball, diff and dsc file for the curl package in the working directory, no browser involved. ^__^;
I use the tracker site when I need to build some packages from unstable or experimental, usually recent libraries that I couldn't find on backports.debian.org.