[llvm-dev] RFC: Introducing an LLVM Community Code of Conduct

Joachim Durchholz via llvm-dev llvm-dev at lists.llvm.org
Wed Oct 14 01:33:22 PDT 2015

Am 13.10.2015 um 22:22 schrieb Renato Golin via llvm-dev:
> This is in no way representative, but it explains why highly technical
> communities have such strong and diverse negative effects, reinforced
> by confirmation bias, mob mentality, etc.

Actually non-tech communities tend to be much worse than technical ones.

Social betterment movements are particularly prone to that - be it 
political parties, charity groups, or prayer circles of whatever 
denomination. The LKML is utterly harmless compared to what's happening 
there - I guess the LKML sparks such interest mostly because of the 
cognitive dissonance, because people expect techies to be more reasonable.
In practice I found they tend to be, actually. Having unquestionable 
facts (the code) and being able to do accurate measurements ("does it 
crash or not", very easy to determine) helps a LOT. Engineering social 
interaction is far more difficult - given that this thread is about such 
a thing, it's no surprise that people jump in.

> This is *not*, however, an argument against having a CoC. A simple CoC
> like "be excellent to each other" is jokingly sufficient, as much as
> listing every possible gender-related variations under the sun on a
> specific state of a specific country is over-specifying and food for
> abuse, as many have mentioned. Neither should be the final document,
> but we haven't struck the balance yet.


> I know this is an RFC, and I think Chandler did the right thing in
> bringing this to the community's attention *before* it was finalised.


> This has already started a lot better than the LLVM Foundation.

Actually I think Tanya explained really well what happened and why.
I'm not worried about the Foundation anymore - not if she can follow up 
with her plan to publish all the stuff that needs publishing.

> 1. It lists too many specific behaviour, mostly only understood by a
> small percentage of the community. Talking about segregating, that
> document is the mother of all segregations. My personal opinion, and
> others seem to agree, we must get rid of every specific detail. A good
> start is the concise list FreeBSD has, ie "Discrimination based on
> gender, race, nationality, sexuality, religion, age or physical
> disability." Though, I'd replace "race" with "ethnicity".

"Discrimination based on things people couldn't change even if they 
wanted". I think that's the core of the anti-discriminatory lists. (It's 
difficult to give that a solid wording though, people will claim "I 
cannot change that I'm rude", but you still want to expel the rude ones.)

> 2. It gives too much power to a group that doesn't exist yet, nor has
> any hint as to how that group would be formed.

There has been a clear statement that the committee is supposed to be 
elected from the community, so unless that statement is retracted this 
issue is off the table.

I think we should try and restrict ourselves to be more on-topic here, 
the subthreads did start to meander.
Apologies for not being fully compliant with that myself. Like you, I'm 
bad at writing short mails...


More information about the llvm-dev mailing list