Re: Rename offensive terminology (master)

From: Andrew Ardill <andrew.ardill@gmail.com>
Date: Jun 15, 2020, 05:19 UTC
Message-ID: <CAH5451nUVB33RkSM2nzJ_yiKF_zSnWGj1FzHwMetg_yVUH05Cw@mail.gmail.com>
In-Reply-To: <20200615044533.GA24311@sigkill.com>

Hi J. Paul, who would have thought! :)

On Mon, 15 Jun 2020 at 14:45, J. Paul Reed <preed@sigkill.com> wrote:

> Heya Andrew... turns out I read this list too, so... thanks for referencin'
> all my work!

No problem, I appreciated the effort you went to.

Show 5 quoted lines

> > Even if git borrowed 'master' from BitKeeper AND BitKeeper used it in
> > a "Master and Slave" fashion, that doesn't mean that the person
> > introducing it to git was using it in a "Master and Slave" fashion,
>
> https://marc.info/?l=git&m=111968031816936&w=2

This is an example we discussed on twitter, so I'll paraphrase/extend on what I said there.

This is definitely a "Master and Slave" usage by Linus, but is not an example of using "Master and Slave" terminology with respect to git branches. It is talking about mirroring kernel.org from a master to the mirror (slave).

It's evidence that Linus used "Master and Slave", but not (to my mind) evidence that the master branch in git was named master because of the "Master and Slave" meaning of master. It may have been, but this example doesn't seem like evidence for that.

> https://marc.info/?l=git&m=111634468526506&w=2

I hadn't seen this one before, but this doesn't seem directly related to this issue either.

Linus says that "the public stuff [note - he is referring to master.kernel.org:/pub/scm/linux/kernel/git/torvalds/linux-2.6.git] is the _slave_. It has no meaning. The only important one is the one that the _developer_ works on."

He is saying that the public master is a slave to (as in a mirror that follows) his local repository. This is definitely "Master and Slave" usage (though an unusual one, as something called master is acting like a slave) but again it is not referring to how branches are named.

In that email, most of his usage is talking about "central repositories" or "main repositories", and "local repositories" which are referred to as workspaces; the only reference to slave is the behaviour of his public repository, in that it follows his local.

Show 6 quoted lines

> Oops.
>
> > It's just as likely that the 'master' usage was common in the industry
>
> Do you have any specific references to, specifically, common usage in the
> industry, at that time?

Nothing apart from those in my previous email - all of the BitKeeper references, and the book which does explicitly use the "Master and Slave" terminology - https://www.google.com/books/edition/Open_Sources_2_0/q9GnNrq3e5EC?hl=en&gbpv=1&pg=PA29

I wasn't focused on what general usage of master may have been, so didn't search out other examples, but the reason I think it's a reasonable alternative are all the usages from BitKeeper (some, such as the airgap example are very old, relatively) and the fact that so many developers today seem to think the usage is the "Master Copy" meaning (and assuming some level of continuity in language over the last 15 years).

In any case, this is not a strongly held belief of mine, my starting point was that most people (myself included) seem to assume the master default branch in git is the "Master Copy", and I haven't yet seen much evidence to suggest otherwise. Happy to see other evidence, and I'll go looking for some myself (in my obviously copious free time haha!)

Show 14 quoted lines

> > My conclusion?
> >
> > Of all the usages of master in BitKeeper, the overwhelming majority of
> > them are of the "Master Copy" variant, consistent with how I and many
> > other people I have seen comment understand gits usage of the term
> > master.
>
> See above.
>
> > To reiterate my point at the top - I believe this information is
> > irrelevant when deciding what git should do now, and my preference
> > would be to have no default at all.
>
> Cool. Sounds like we mostly agree...

Cool indeed :)

Regards,

Andrew Ardill