LWN: Comments on "Who wrote 2.6.32"

Still no Canonical?

Cato — Wed, 09 Dec 2009 14:18:07 +0000

Ignoring Canonical's major contributions in usability and integration above the kernel is also pretty "sad", not to say biased. If you weren't working for a competitor, your harping on about this would be more credible - I prefer Red Hat's approach in just developing great new features in Fedora.

Developer credit

corbet — Mon, 30 Nov 2009 14:16:44 +0000

It's taken from the "author" credit in git. Subsystem maintainers generally take pain to get that right, so it's a pretty reliable metric. I only know of a couple of significant exceptions; one was in the early days of the staging tree, and the article at that time explained the issue. I did also make a correction for one misattribution in the 2.6.32 series.

Who wrote 2.6.32

broonie — Mon, 30 Nov 2009 11:04:11 +0000

It's From: AFAICT.

Who wrote 2.6.32

broonie — Fri, 27 Nov 2009 14:49:57 +0000

While a good proportion of my code (more than average this time around) is device support there's also a good proportion of subsystem work and CPU side support in there too (hence also the appearance in the non-author signoffs).

Google and 2.6.32

alex — Fri, 27 Nov 2009 13:14:19 +0000

I suspect the board support stuff came in big chunks without the development
history behind them, hence the low changeset count.

Who wrote 2.6.32

rvfh — Fri, 27 Nov 2009 12:43:10 +0000

Indeed, that seems odd as we always complain about their attitude regarding wireless drivers... I assume they are more active in the wired side of things.

Google and 2.6.32

mbanck — Thu, 26 Nov 2009 15:22:57 +0000

OK, I indeed missed it. They contributed 40k lines, but only a few changesets.

Sorry.

Google and 2.6.32

mbanck — Thu, 26 Nov 2009 14:00:34 +0000

Google takes a lot of grief for not contributing back, but that company was the source of a fair amount of code going into 2.6.32.

I don't see Google in the above statistics, were their "fair amount of code" below the cutoff, or was this a typo and Google contributed to 2.6.31 (or 2.6.33)?

Drop staging from the statistics?

wsa — Thu, 26 Nov 2009 09:17:23 +0000

I agree and would again vote for excluding staging from these statistics (or make it a seperate list). Most people here probably know how to read these rankings, but most not involved in kernel development won't. IMHO it's just too easy to make 'big headlines' out of nothing this way. The ultimate nightmare would be some "Let's dump a driver and get famous"-tourism getting popular within marketing areas ;)

Still no class?

foom — Thu, 26 Nov 2009 05:12:16 +0000

> I think that is just petty tbh.

I went and looked up the linked article, an official blog by Novell's Chief Marketing Officer:

Of course this announcement is about much more than 20,000 lines of code Microsoft is committing (which by the way once accepted into the Linux tree will far surpass those contributed by Canonical).

Wow, yes, that really does seem petty.

Microsoft's contribution is just a driver for improving the speed of linux guests hosted on their proprietary virtualization platform. I'm sure it's nice for Windows users, but it doesn't matter whether that's 20 kLOC or 2 LOC: it's still an isolated driver for virtualization on top of proprietary OS. And I have no doubt that it helps further Windows sales and deployment way more than it helps further Linux deployment...

I'm certain that if Canonical has contributed even one line of code to Linux, it will have been a more valuable contribution, to me, than the 20kLOC that Microsoft contributed. (...which they only did under duress, I'm led to believe).

That said, it would of course be wonderful for Canonical to do more kernel devopment...

Who wrote 2.6.32

mcisely — Thu, 26 Nov 2009 05:07:24 +0000

Is that first table "Most active 2.6.32 developers" really tallied using changeset "from:" headers or is it "signed-off-by:" headers? If it is signed-off-by then at least in one case I think you are crediting the subsystem maintainer for all the work done by those actually working on that subsystem. Just a guess.

Also, something strange: Mauro is credited with 62719 lines of code changed yet that is still not enough changesets to break into the top-19 list by changeset? That means that those 62719 lines of code are spread over less than 70 changesets. That's nearly 900 lines / changeset. Seems out of scale. Maybe it was skewed by just a few enormous cases?

Still no class?

kragil — Thu, 26 Nov 2009 02:12:48 +0000

I think that is just petty tbh.

So MS dropped a big pile of code to improve Linux as a guest on their platform.
Canonical just fixed a few real problem linux users had.

I'll prefer Canonicals contribution to MS' any day.

Only when Linux has reached world domination we can really tell who contributed most to its success.

I wouldn't bet on Novell to win that prize though.

Who wrote 2.6.32

ajb — Wed, 25 Nov 2009 22:15:00 +0000

It's also notable that Broadcom was 4th in terms of lines changed, after None, Novell, and Red Hat.

Still no Canonical?

gregkh — Wed, 25 Nov 2009 22:08:19 +0000

No, I have not run any stats on stuff higher up than the plumbing, as I'm personally not interested in those ecosystems at this moment.

But all of the scripts that we use to generate this information are published, if anyone else wants to do it, feel free to do so.

Still no Canonical?

rahulsundaram — Wed, 25 Nov 2009 22:05:19 +0000

A bit more at

https://www.redhat.com/archives/fedora-devel-list/2009-No...

Still no Canonical?

rahulsundaram — Wed, 25 Nov 2009 22:03:54 +0000

Although not directly related to your area of interest, I wonder if you have run detailed stats on things a bit higher up the stack including desktop environments like GNOME and KDE? I am curious to know how different the picture looks. Since you went to the level of Xorg, I don't think its a strech to look more holistically at a enlarged core Linux ecosystem.

Still no Canonical?

gregkh — Wed, 25 Nov 2009 22:03:05 +0000

Already fixed and in the linux-next tree. See: http://lkml.org/lkml/2009/11/11/336 for details.

Still no Canonical?

hppnq — Wed, 25 Nov 2009 22:00:13 +0000

Well, here's something to pick you up!

memset(request, sizeof(struct storvsc_request_extension), 0);

Industrial-strength hilarious code is exactly what this kernel lacked all these years.

Still no Canonical?

gregkh — Wed, 25 Nov 2009 21:22:26 +0000

Yes, it's sad:
http://identi.ca/notice/6753919

Still no Canonical?

nevets — Wed, 25 Nov 2009 18:44:08 +0000

OK, I have to say it. I never expected Microsoft to show up on that list before Canonical!