Is the master branch in AUR mandatory?

Chispy · 2020-07-07 17:47:26

Hi,

I'm the maintainer of https://aur.archlinux.org/packages/tinygo-bin/, and we're making an effort in the community to move away from old master-slave terminology. To that end, I'm wondering if I can use another branch in the AUR git repos? Would it work if I delete the master branch and leave a release branch alive as the only remote branch?

Thanks!

eschwartz · 2020-07-08 01:02:02

Yes it is mandatory. See the aurweb backend code for details.

Fortunately, the word "master" isn't inherently evil, and has meanings unrelated to the word "slave" -- for instance, the git usage of it.

ewaller · 2020-07-08 01:29:13

And, now that we have our first response to this thread, please allow a moderator comment.

eschwartz, thank you and I agree. And, I resist attempts to pare down language. 1984 was meant as a warning, not a guide.

Regardless, recent events have reinforced the need for inclusiveness and the need to eliminate non-inclusive language. Master-slave does has negative connotations. In my personal opinion, Master craftsman does not. Neither does Master of Science. On the other hand, other's may disagree; maybe there are issues with the master-apprentice relationship historically.

None-the-less, This thread touches on recent trends in the software community at large, and we are not going to argue the point here; there are plenty of other places to do so. We are going be respect each other, and we are not going to allow politics to be argued here.

Moderators will ensure this conversation stays productive and civil.

Chispy · 2020-07-08 11:41:11

Note that while the origin of the name for the default branch being "master" in git is not clear either way, I don't think it can be asserted that it has nothing to do with the master/slave terminology. See the discussion on the gnome mailing list for more information: https://mail.gnome.org/archives/desktop … 00066.html.

In any case, thanks for answering the question, which I suppose leads me to ask if we would consider making it configurable? I'd be happy to make this contribution myself if it's available open source somewhere.

Thanks!

Slithery · 2020-07-08 11:49:30

Chispy wrote:

...if it's available open source somewhere.

https://git.archlinux.org/aurweb.git/

mpan · 2020-07-08 11:53:41

The purpose of VCS entries in AUR is to represent the tip of the main branch.⁽¹⁾ In Git the main branch is called “master”.⁽²⁾ QED
____
⁽¹⁾ VCS package guidelines
⁽²⁾ Git Book, Glossary

Trilby · 2020-07-08 12:26:59

mpan, your linked reference doesn't support your claim. The git glossary does not specify that "master" is the main branch and even explicitly notes it is not even required. I believe a claim that it is a default branch could be supported by that link, but not the main branch. This is coded into the AUR workings as linked above. Code can be changed.

Personally I don't think it should be changed - for reasons that may not be fitting to discuss here. But if the question is whether it's possible to change it, it certainly is.

Last edited by Trilby (2020-07-08 12:27:17)

mpan · 2020-07-08 13:22:11

My goal was to avoid any discussion in this thread, but it seems I have no choice.

It is the default branch and by convention it is used as the main one. Yes, the administrator may choose any branch for that purpose, but this is what is the default and what pretty much everyone uses, which is reflected in the linked paragraph.

Trilby · 2020-07-08 13:26:12

In what way is it reflected in that paragraph? Show me where it says that "master" should be the main branch. It doesn't. What it does say is the name "is purely by convention and is not required."

If you didn't link to that glossary and simply said that keeping the name "master" is what a vast majority of people do - I'd agree. Though I'm not sure how relevant that is - the purpose of this thread is to suggest a change in the most common behavior, so countering by pointing out that it is indeed the most common behavior is rather pointless.

But you didn't just say it was the most commonly used branch name, you said linked to that glossary as evidence that "master" should be the main branch. And that is simply not supported by the glossary entry. The main branch is not by design or requirement called "Master", the default branch is, not the main branch.

Last edited by Trilby (2020-07-08 13:32:13)

Awebb · 2020-07-08 13:54:57

This is the wrong place to discuss this, anyway. All development happens on the mailing list and maybe on the bug tracker. This is the support bbs, not the right place for a feature request.

EDIT:

There is no issue with a master-apprentice relationship. This is still the default education path in Germany. This whole newspeak nonsense comes up every few years and leaves its mark on the usability or language. I recently booked a conference room, because somebody was unable to hyphenate a double name and then neglected to indicate, that "they" is this new pluralis majestatis third person and not an actual group of people. Wowzer.

Last edited by Awebb (2020-07-08 14:01:27)

Trilby · 2020-07-08 14:04:17

Awebb wrote:

There is no issue with a master-apprentice relationship. This is still the default education path in Germany.

And when we think cultural sensitivity and race relations, we think Germany, right?

schard · 2020-07-08 14:28:16

Well, yes the master/slave terminology is certainly mainly an issue in English speaking countries with a certain history.
Also @Awebb's analogy is misleading. The terms are Meister and Geselle. And these are not related to historic slavery in the US.
The same word, when translated, can have different connotations in different languages.
I always feel a bit guilty, when I investigate the socket statistics on my systems.

As for the original discussion: What would be the benefit from allowing additional branches in the AUR code.
It just adds complexity. And I suspect that many maintainers won't branch all their projects just to rename the branch to something that is supposedly less offensive.

Last edited by schard (2020-07-08 14:29:57)

Awebb · 2020-07-08 15:05:03

Trilby wrote:

Awebb wrote:
There is no issue with a master-apprentice relationship. This is still the default education path in Germany.
And when we think cultural sensitivity and race relations, we think Germany, right?

That would be really funny, if it wasn't so ironically sad... or sadly ironic.

eschwartz · 2020-07-08 16:48:42

Chispy wrote:

Note that while the origin of the name for the default branch being "master" in git is not clear either way, I don't think it can be asserted that it has nothing to do with the master/slave terminology. See the discussion on the gnome mailing list for more information: https://mail.gnome.org/archives/desktop … 00066.html.

So it's "not clear" whether the origin of the name has to do with the master/slave terminology, yet at the same time it has to do with it regardless?

I can, in fact, assert it.

git is not bitkeeper
bitkeeper is not influential or well-known
the person who named git "master" asserts having done so by independently coming up with the name, and not realizing it had an alternative meaning connected to slavery -- either he's lying through his teeth, or it isn't a slavery reference
git has no concept of slavery, all the more so since git is all about emotionally validating your right to fork or branch, and the goal of forking is to take over mindshare and become the official "master" branch everyone considers to be the master reference, while the goal of branching is to eventually merge your branch into master, at which point the new master is a union of the old master + the old topic branch (or more often, the new master IS the old topic branch, fast-forward merges FTW)

I'll add to this, that just following your link onward to the bitkeeper links:

bitkeeper documentation indicates bitkeeper has no concept of branches, but refers to "master-directory" as the directory name of a remote
different bitkeeper documentation refers to master.dmn.com and slave.dmn.com as *server hostnames* and a repository named /u01/my_package
I can absolutely believe bitkeeper is using slavery terms, but I also believe this metaphor is so forced, the bitkeeper developer writing this documentation was so unbelievably high on LSD as to set a local record deeply, deeply confused and rambling
I doubt even people who used bitkeeper were well aware of the documentation's use of example server hostnames, and if I were a bitkeeper user I can fairly well guarantee I would not edit /etc/hostname on my laptop to read "slave" just because I once ran bitkeeper on it. I most likely would not use master.archlinux.org to push code either, though I might use bitkeeper.archlinux.org; instead I use git.archlinux.org (a role name which is interchangeable with luna.archlinux.org as the physical name)
I'm essentially positive people who didn't really know about bitkeeper except maybe having heard the name and copy/pasted some commands to download source code, have/had no awareness of some documentation-only references that don't even make coherent reading
based on the git developer's testimony re: not realizing the alternative slavery-related meaning, it seems fairly obvious that this individual either never heard of bitkeeper, or only knew about bitkeeper through copy-pasting a handful of commands listed on kernel.org and did not ever read LSD-influenced forced metaphor documentation

Hence I believe there's no valid claim that git was in any way even indirectly inspired by bitkeeper. Correlation does not imply causation, and this is a pretty weak correlation at best.

Finishing notes:

The gnome mailing list is frequented by people who develop gnome. Given the fact that Gnome's GLib even exists, and given the further testimony of Gnome 3... my opinions about the common sense of gnome developers would probably violate a forum rule. Let's leave it at "I would not ever expect them to do proper investigation".

Chispy wrote:

In any case, thanks for answering the question, which I suppose leads me to ask if we would consider making it configurable? I'd be happy to make this contribution myself if it's available open source somewhere.
Thanks!

Linux distros generally are open source, yes... someone else already linked you to it.

To elaborate on my previous statement, the aurweb backend internals remap the branch ("master") you push into their internal storage ("$pkgname") and which is visible on the single-repo cgit interface. The AUR does not support multiple branches, because there's no valid use case (it is a publishing platform backed by git, not a development forge) and *theoretically it could use any name for mapping*. The one chosen was the one which newbie git users got when following the instructions to 'git init && git add PKGBUILD && git commit && && git remote add origin aur@aur.archlinux.org/pkgbase.git && git push'.

However, we will NOT break existing clones for 60,000+ packages, this is non-negotiable, so the logical conclusion is, those MUST continue to expose their single-branch history as "master".

As for setting a different init-time name, it's not an automatic non-starter, but... how would this work exactly? We need to map something predictable, and it should not break the workflow of users who run 'git init && git add PKGBUILD && git commit && git push'. It should also not break my aurpublish tool, which uses git subtree and a single repository to reproducibly spin out packages and push them to "aur@aur.archlinux.org/$pkgbase.git master"; there are no permanent branches, how is the tool supposed to know which branch to push itself as, if the rules change depending on when the AUR date of submission is? It cannot even use git commit dates as those aren't connected to the submission date.

This sounds like a lot of undefined complexity for something which is not even a problem as far as I can tell. And again, we don't operate a development forge, so "support my personal branch naming model" isn't, on its own, a valid argument to add burdensome complexity -- if the permitted branch name were "release", we would never add a configuration option to let users decide on their choice of "trunk" (svn) or "default" (mercurial) or "develop" (git-flow model doesn't even make sense with only one branch) or "archlinux-goes-brrr" (vanity name).

Whatever the name is, it should be unambiguous.

Last edited by eschwartz (2020-07-08 16:55:18)

Head_on_a_Stick · 2020-07-08 20:11:32

Can I leave this here without irritating everybody immensely?

https://ferd.ca/inclusiveness-in-langua … ng-in.html

I think Fred explains the situation quite well.

Roken · 2020-07-08 22:33:15

I'm going to add dictionary definitions (not all of them. As with many words in the English language, there are varied meanings)

Master~: main; principal.

Slave~:subject (a device) to control by another.

In terms of software, I see no issue with either. It's all about context.

Last edited by Roken (2020-07-08 22:38:12)

jasonwryan · 2020-07-08 22:43:58

Roken wrote:

I'm going to add nothing to this discussion by selecting only the definitions that suit me.

Slithery · 2020-07-08 23:11:05

OK folks, some of these posts are getting too political for the scope of this forum.

Please stick to discussing technical details only, otherwise this thread will have to be closed.

I've read hundreds of posts on different forums recently about this topic and they've all descended into dark pits of chaos. Hopefully the Arch community can provide a shining light into how things should be discussed in a mature fashion.

Slithery.

Trilby · 2020-07-08 23:20:21

Should the thread then be closed? I think Eschwartz's post thoroughly covers the technical hurdles (understatement) to implementing this. I'm not sure how much more could be said on a technical end to address the question of whether this change can be made. That only leaves discussion of the reasons to want it to change ... which is all political.

Last edited by Trilby (2020-07-08 23:20:43)

fukawi2 · 2020-07-09 00:36:27

Yep, we're done here. The OP question has been answered, subsequent political discussion only serves to be divisive.

Closing.

Arch Linux

#1 2020-07-07 17:47:26

Is the master branch in AUR mandatory?

#2 2020-07-08 01:02:02

Re: Is the master branch in AUR mandatory?

#3 2020-07-08 01:29:13

Re: Is the master branch in AUR mandatory?

#4 2020-07-08 11:41:11

Re: Is the master branch in AUR mandatory?

#5 2020-07-08 11:49:30

Re: Is the master branch in AUR mandatory?

#6 2020-07-08 11:53:41

Re: Is the master branch in AUR mandatory?

#7 2020-07-08 12:26:59

Re: Is the master branch in AUR mandatory?

#8 2020-07-08 13:22:11

Re: Is the master branch in AUR mandatory?

#9 2020-07-08 13:26:12

Re: Is the master branch in AUR mandatory?

#10 2020-07-08 13:54:57

Re: Is the master branch in AUR mandatory?

#11 2020-07-08 14:04:17

Re: Is the master branch in AUR mandatory?

#12 2020-07-08 14:28:16

Re: Is the master branch in AUR mandatory?

#13 2020-07-08 15:05:03

Re: Is the master branch in AUR mandatory?

#14 2020-07-08 16:48:42

Re: Is the master branch in AUR mandatory?

#15 2020-07-08 20:11:32

Re: Is the master branch in AUR mandatory?

#16 2020-07-08 22:33:15

Re: Is the master branch in AUR mandatory?

#17 2020-07-08 22:43:58

Re: Is the master branch in AUR mandatory?

#18 2020-07-08 23:11:05

Re: Is the master branch in AUR mandatory?

#19 2020-07-08 23:20:21

Re: Is the master branch in AUR mandatory?

#20 2020-07-09 00:36:27

Re: Is the master branch in AUR mandatory?

Board footer