LWN: Comments on "Looking back at the UMN episode"

Looking back at the UMN episode

pabs — Tue, 05 Oct 2021 02:57:26 +0000

> I've been told [contributing to Linux mainline] is not the kind of thing the [Linux] Foundation is interested in encouraging.

Uh, that seems very strange, did they give any reasons for that?

Looking back at the UMN episode

ghane — Mon, 04 Oct 2021 01:38:28 +0000

GKH: "...he gets to be mad once per year, he said, and this was the time for this year. "

So I can start sending in my patches? I should be safe till Christmas :-)

Looking back at the UMN episode

tedd — Sat, 02 Oct 2021 08:01:47 +0000

With your and the kernel's experiences, why the hell would you still give UMN any attention at all? I know I won't have any time for the bullshit they pull.

Looking back at the UMN episode

marcH — Thu, 30 Sep 2021 16:29:33 +0000

Great summary, unfortunately not specific to the US.

In big corporations the incentives are wrong too: to get promoted you need to produce something that looks "ground-breaking" too whether it actually is or not. Hard and patient work that merely pleases the customers is too difficult to measure and it does not get you very far either HOWEVER it still pays your bills in most places (just avoid the next Theranos).

1% inspiration and 99% perspiration gets you absolutely nowhere in academia, I mean at least not in computer science, I think other fields tend to be better. If you're young and reading this then don't go there ever, that system is broken by design. If you really want to try something outlandish then find some startup instead.

Looking back at the UMN episode

deater — Thu, 30 Sep 2021 15:40:13 +0000

Current US academia, especially in the computer area, is fairly broken, and I say that as someone deeply involved in it.

Researchers are judged on grant money, with a side focus on high-profile publications, and that's about it. The stakes are high and sadly there's a lot of politics and cheating involved that's often overlooked because the majority of people involved are afraid the whole scheme will collapse if it's investigated too thoroughly.

It turns out doing actual proper Linux or open-source research that involves working closely with upstream is hard. In addition, making slow gradual improvements to existing code is considered "incremental" and will not get publications or funding.

Wheras hiring a few low-level students to grab a 10-year old RHEL kernel, fling some poorly-written benchmarks at it, then write up some over-the-top article about "Linux is terrible" seems to be a winning way to get a top publication. Things are peer-reviewed, but understand it's generally other similar minded professors doing the peer-review (or even their grad students if the big name person is "too busy") rather than knowledgeable people from industry/open-source.

The articles will often propose preposterous "solutions" to the problems they find (as per the original UMN work. Similarly, I've been at PhD defences where O(N^3) algorithms were proposed for the scheduler and none of the experts batted an eye).

They don't like fixing bugs either because their tools look less impressive if they can't claim 1000 bugs found anymore because they fixed things upstream.

There really isn't a good solution to this in the current environment. I've suggested in the past that maybe the Linux Foundation could offer grants, sort of like google-summer-of-code, but for longer (3 year?) terms with strict wording requiring proper contributions back. I've been told this is not the kind of thing the foundation is interested in encouraging.

Looking back at the UMN episode

NYKevin — Thu, 30 Sep 2021 01:33:51 +0000

Academics are used to the idea that they can just get a few smart people in a room, have them look at a technical issue, and the consensus of that group will be The Right Answer, regardless of what anyone else thinks. This is how they write and grade exams, after all.

Looking back at the UMN episode

mathstuf — Tue, 28 Sep 2021 20:37:57 +0000

There is a program[1] at my alma mater that worked out pretty well (at least it was when I was there and for years after that, though the main profs behind it have since retired). The associated class had an assignment to engage with an existing project and contribute in some way (I chose Wesnoth and got a few patches in). Later, I went back after graduating to help out with a Linux kernel patch submission and even got one in myself (cleaning up a long-neglected network driver whitespace and `printk` usage pattern updates). I had been doing KDE and Fedora work already by that point, so I wasn't new to it, but I think it helped many at least interact with some communities out there.

I will say that this is by far focused on students rather than researchers, but professors are engaged with the program. I just don't know how relevant it is to their research.

[1]https://rcos.io/

Looking back at the UMN episode

olof — Tue, 28 Sep 2021 19:19:16 +0000

Is there a success story to highlight of a university system investing anything like a mini-opensource-office for helping their researchers (and students?) engage with outside opensource communities?

Seems like it's something the corporate world has figured out the benefit of, but it sometimes takes some sort of engagement failure for them to realize the need to invest there.

Looking back at the UMN episode

error27 — Tue, 28 Sep 2021 06:02:51 +0000

I'm still annoyed at the UMN devs because they blamed me for not catching their "bug" even after I emailed them.

Them: Ha ha. You caught some bugs in our patch but you didn't catch our use after free!
Me: It's not a use after free. The caller is holding a reference.
Them: Well, we still think it was a use after free.

Ah well...