LWN: Comments on "Some development statistics for 6.2"

Some development statistics for 6.2

nix — Tue, 28 Feb 2023 19:18:04 +0000

Hm. gitglossary(7) says:

> BitKeeper/cvsps speak for "commit". Since Git does not store changes, but states, it really does not make sense to use the term "changesets" with Git.

(I wonder if the persistence of the term here is just because Jon spent so much time reporting on Linux in the almost-forgotten BitKeeper days. On the other hand, in one month git will be old enough to vote...)

What happens to new contributors

sima — Mon, 27 Feb 2023 16:46:02 +0000

I guess you could fix that by only peeking a constant amount of time into the future for every release. And then maybe for the oldest release you look at make a statistics of how many people returned that your arbitrary cut-off (maybe 1y or so) counted as lost?

Still a pile of scripting work to generate those numbers ...

What happens to new contributors

Nikratio — Fri, 24 Feb 2023 09:15:58 +0000

You are completely right. I do not see why this makes the data less interesting though :-).

Another interesting thing may be to plot a histogram of current developers vs number of past kernels they have contributed to. This would tell us more about how much of the kernel comes from regular contributors vs occasional ones.

What happens to new contributors

Wol — Wed, 22 Feb 2023 16:45:36 +0000

Maybe stats to say how many contributors have contributed to how many kernels?

Dunno how far you want to go back, but maybe multiply the "contributors to the current kernel" by say 4, and then provide stats for the most prolific "contributed to X kernels" contributors up to that number.

Yes that will lose the "drive by" contributors, but it'll give an insight into the "now and then" contributors. Ten years is probably a good time window.

Cheers,
Wol

What happens to new contributors

corbet — Wed, 22 Feb 2023 16:00:26 +0000

The number of contributors to each release does continue to grow.

Meanwhile, I have at times looked at lost contributors — those who contributed to a given release for the last time. The problem is that such a signal is necessarily old; we do get people who show up every few years to fix something that bothers them. You can never really say that somebody is gone.

Some development statistics for 6.2

Nikratio — Wed, 22 Feb 2023 15:35:49 +0000

Here's an idea for the next such article: maybe you could look at how the population of kernel developers changes over time?

If for a given kernel there's hundreds of new first-time contributors, what happens to them afterwards?

The overall number of contributors does not seem to be changing that much. So are the new contributors replacing older contributors, or do they never come back after the first patch?

Some development statistics for 6.2

Wol — Wed, 22 Feb 2023 12:37:49 +0000

> but I have the feeling that changeset is not a term commonly used, or is it?

Changeset is jargon.

In other words, the man in the street won't have a clue, but it's used by programmers, and your typical programmer will understand it.

A lot of people moan about jargon, but at the end of the day all jargon is is an industry-specific vocabulary. You cannot expect people with a shared experience to talk about that experience using only words in widespread usage. They will use words specific to that experience, and that's what jargon is. (And it pisses me off greatly, when I use correct, accurate words and other people moan because they don't understand them and want me to explain in twenty words when one should do.)

Cheers,
Wol

Some development statistics for 6.2

jonas.bonn — Wed, 22 Feb 2023 11:15:11 +0000

A commit is the application of a changeset to branch's linear history. A changeset can be committed to several different branches/repos. Given that this article refers to the main branch of Linus' repo, in this case there's a one-to-one relationship between a commit and a changeset, but in general I'd say developers create changesets and maintainers create commits thereof.

Some development statistics for 6.2

jtepe — Wed, 22 Feb 2023 08:12:06 +0000

I always wondered, why call it changeset instead of commit? From my understanding there is no difference here. I'm not a native English speaker but I have the feeling that changeset is not a term commonly used, or is it?

Some development statistics for 6.2

calumapplepie — Tue, 21 Feb 2023 21:05:21 +0000

Deleted a number of unmaintained drivers?

Gimme a sec, I need to go buy some popcorn for when the inevitable crew of people shows up complaining about broken workflows and demanding that the drivers be restored /j. Also possibly look for a list of removed drivers to see if I should put down the popcorn and join them.

Some development statistics for 6.2

pbonzini — Tue, 21 Feb 2023 17:04:56 +0000

I checked a couple old statistics articles (4.0 / 4.5 / 4.10) and it's interesting that while IBM contributions did decline quite a bit, Red Hat's have actually gone up in terms of number of commits. It's Google and Linaro that have grown a lot more and stole the third and fourth place, like Intel had done during the 3.x days.

Some development statistics for 6.2

corbet — Tue, 21 Feb 2023 14:39:32 +0000

Well, it averages to 6.2 :)

The two weird typos have been fixed; the actual data was correct and has not been changed.

Some development statistics for 6.2

tpetazzoni — Tue, 21 Feb 2023 10:51:37 +0000

It seems like there is an issue with the employers list. The text above says " A total of 235 employers supported work on 6.3, a fairly normal sort of number", and then the legend of the table says "Most active 6.1 employers". Shouldn't both refer to Linux 6.2 ? It also seems like the table contents themselves are not up to date with Linux 6.2.

Some development statistics for 6.2

Kamiccolo — Tue, 21 Feb 2023 00:28:43 +0000

Wihihi, adorable graphical addition! Thanks a lot!