Re: [arch-dev-public] rc.d files, unmaintained packages - and, the quality of our repositories
On 16/07/13 04:39 AM, Gaetan Bisson wrote:
However, now netcfg has been readded to community in the exact same state as the package that was originally removed:
* It does not work properly with systemd. * There is no init system in our repositories that it works with. * It actually re-added rc.d files to our repositories, although we had a TODO list recently to explicitly remove those (and btw, they depend on files that no longer exist in our repositories, like /etc/rc.d/functions and /etc/rc.conf).
I am really confused about the decision to re-add this and I am seriously considering if we should talk about stricter guidelines for adding packages and - in particular - the quality of our packages. I doubt guidelines would help. It should be pretty obvious to any responsible packager that re-adding a deprecated package violating recent TODO lists is a bad idea. If we really need to spell this out (with an exhaustive list of obvious things responsible developers should not do), then we have bigger problems.
In this particular case, we should hear what Connor has to say and make sure (one way or another) that this type of problem will not happen again.
Did you see my thread in arch-projects? It's probably safe to assume other TUs have more sense than me.
[2013-07-16 10:40:24 -0700] Connor Behan:
Did you see my thread in arch-projects?
Your message there does not say why you think adding netcfg back to the official repositories was a good idea; it merely states that you did so. Besides, arch-projects is the wrong list to discuss our repositories. -- Gaetan
On 16/07/13 04:08 PM, Gaetan Bisson wrote:
[2013-07-16 10:40:24 -0700] Connor Behan:
Did you see my thread in arch-projects? Your message there does not say why you think adding netcfg back to the official repositories was a good idea; it merely states that you did so. Besides, arch-projects is the wrong list to discuss our repositories.
I have stopped using digests for this list now, so here's a better reply. There are a few packages I can name and a few I can't that assume the user runs systemd. So this makes it is a bad idea to upload a different init system any time soon. However, I don't think there is anything that assumes the user has netctl. As long as someone steps up to develop it (which Florian seemed to encourage), netcfg should be no worse an alternative than wicd or networkmanager. I used systemd+netcfg on a server until this month and there were no issues. Even so, I put the word "predecessor" in the pkgdesc as a warning. Yes it has bugs, but so do a lot of [community] packages and the criteria for inclusion on the wiki said "1% usage from pkgstats or 10 votes on the AUR." Pkgstats would obviously be unreliable because of people who are slow to update so I waited for the package to get 10 votes. My rush to release it last night was a different story. I should've removed the rc.d files (initscripts users get these from elsewhere now anyways) and I definitely should've removed the base group. In case anyone took my request on arch-projects seriously, feel free to reject it now that you know it came from a klutz. But honestly, netcfg was one Arch project that was actually useful outside of Arch so I still think maintaining it would be a good use of my time. And I thought I would force myself to dive into it by making a release first and having a discussion after. I will be sure to ask first next time if I ever think it's a good idea to release it again.
[2013-07-16 18:08:08 -0700] Connor Behan:
There are a few packages I can name and a few I can't that assume the user runs systemd. So this makes it is a bad idea to upload a different init system any time soon.
Indeed: we made a concerted decision to use systemd as our init system, so systemd can be assumed to be running on every Arch Linux system. If you wish to go against this decision (such as to support other init systems) you need to submit a proposal here so we can discuss it first. It's really the opposite of "push first and discuss later".
Yes it has bugs, but so do a lot of [community] packages and the criteria for inclusion on the wiki said "1% usage from pkgstats or 10 votes on the AUR."
That criteria is nothing more than a guideline. The concerted decisions we make on this list (in particular the deprecation of netcfg) obviously take precedence - and the proper place to disagree would have been: https://mailman.archlinux.org/pipermail/arch-dev-public/2013-April/024774.ht...
But honestly, netcfg was one Arch project that was actually useful outside of Arch
That's not the opinion of most people; see the last two lines of https://mailman.archlinux.org/pipermail/arch-projects/2013-March/003698.html
And I thought I would force myself to dive into it by making a release first and having a discussion after. I will be sure to ask first next time if I ever think it's a good idea to release it again.
When the upstream maintainers of a project find it too messy and switch their development efforts to a cleaner fork, I find it quite naive to "re-release the messy original now, and see about fixing it later"... But of course you are free to develop netcfg (or a fork) and release tarballs of your changes on your personal website. What you did wrong here is push this as a package to our official repositories, going against decisions made on this very list. -- Gaetan
If you wish to go against this decision (such as to support other init systems) you need to submit a proposal here so we can discuss it first. One could argue that having syslog-ng in [extra] supports other init systems in some small way. This does not come at anyone's expense
This may be a dead horse, but a few points. On 16/07/13 07:20 PM, Gaetan Bisson wrote: though. Unfortunately, netcfg would come at the expense of reputation because it is an Arch project. Supporting things deprecated by Arch itself sends a worse message than supporting things that were deprecated by others.
When the upstream maintainers of a project find it too messy and switch their development efforts to a cleaner fork, I find it quite naive to "re-release the messy original now, and see about fixing it later"... Again, I think the key here is that it was an Arch project. If for some reason a TU released community/thunar-thumbnailers "Predecessor to tumbler" most people wouldn't care. A project's own developers are the most likely people to think it is messy.
As for the whole "push first discuss later" thing, people shouldn't be afraid to take this approach. Maybe I misused it but "back in my day" there was an Arch dev who had wise words <https://bbs.archlinux.org/viewtopic.php?pid=355343#p355343> about this sort of thing.
On Wed, Jul 17, 2013 at 9:06 PM, Connor Behan <connor.behan@gmail.com> wrote:
As for the whole "push first discuss later" thing, people shouldn't be afraid to take this approach. Maybe I misused it but "back in my day" there was an Arch dev who had wise words <https://bbs.archlinux.org/viewtopic.php?pid=355343#p355343> about this sort of thing.
You divert completely the spirit of the sentence! Of course we need people who DO, but It not mean you can be out of control. Pushing back a package without the _minimum_ level of quality and review is something people shoud be afraid! Ioni can release the Kraken. By your light way of doing, you cause troubles to users of netctl and your last word is "shoot first, dicuss later, is a not so bad". A better approach would be to fix the issues, send a mail which explain what you already done and annoucing that you want push netcfg back. We all make mistakes, please, try to understand your teammate trying to make you improve your work. Cheers, -- Sébastien "Seblu" Luttringer https://www.seblu.net GPG: 0x2072D77A
On Wed, Jul 17, 2013 at 11:34 PM, Sébastien Luttringer <seblu@seblu.net>wrote:
By your light way of doing, you cause troubles to users of netctl
Not taking a particular side here but maybe I'm missing something because I don't understand why netctl users would be troubled by this. -- Massimiliano Torromeo
On Thu, Jul 18, 2013 at 12:58 AM, Massimiliano Torromeo <massimiliano.torromeo@gmail.com> wrote:
On Wed, Jul 17, 2013 at 11:34 PM, Sébastien Luttringer <seblu@seblu.net>wrote:
By your light way of doing, you cause troubles to users of netctl
Not taking a particular side here but maybe I'm missing something because I don't understand why netctl users would be troubled by this.
I referred to this. https://mailman.archlinux.org/pipermail/arch-general/2013-July/033828.html -- Sébastien "Seblu" Luttringer https://www.seblu.net GPG: 0x2072D77A
[2013-07-17 12:06:24 -0700] Connor Behan:
As for the whole "push first discuss later" thing, people shouldn't be afraid to take this approach. Maybe I misused it but "back in my day" there was an Arch dev who had wise words <https://bbs.archlinux.org/viewtopic.php?pid=355343#p355343> about this sort of thing.
There is quite a big difference between writing code and pushing deprecated code to our repositories; I really hope you can see that. Obviously you are free to improve netcfg's code; but pushing it to our repos is a different story. Could we agree that going against the concerted decisions of official packagers (as made on this list) is obviously wrong, and that you will not do it again in the future? It'd be nice if we could all move on... -- Gaetan
On 17/07/13 04:51 PM, Gaetan Bisson wrote:
Could we agree that going against the concerted decisions of official packagers (as made on this list) is obviously wrong, and that you will not do it again in the future? It'd be nice if we could all move on... Yes, never again.
participants (4)
-
Connor Behan
-
Gaetan Bisson
-
Massimiliano Torromeo
-
Sébastien Luttringer