[llvm-dev] Resuming the discussion of establishing an LLVM code of conduct

Joachim Durchholz via llvm-dev llvm-dev at lists.llvm.org
Thu May 5 17:03:50 PDT 2016

Am 06.05.2016 um 00:06 schrieb Tanya Lattner via llvm-dev:
> The point is that you wouldn’t know that from looking at the
> alternative code of conduct. I would need to spend months pouring
> through mailing lists posts and watching the community to feel its a
> good and safe place.

First, there's no such thing as a "good and safe place" per se; what's 
good and safe for one person might be abusive for another person. 
Different cultural and personal backgrounds, and if you join an 
international community, you actually expect people to be different.

Second, words are cheap. Even an enforceable CoC won't make you feel 
safe, because you'd have to pore over months of mailing list posts to 
see whether it is actually enforced.
The only thing that can be concluded from an enforceable CoC is that it 
may be enforced if you misbehave, in the eyes of somebody with enough 
clout to level that kind of allegation. I'd feel much *less* safe in 
such an environment; so I guess it's good thing that I don't care too 
much about safety but simply participate and watch what happens.

 > One of the many reasons a code of conduct is
> useful is to show outsiders what our community is like, what we think
> is acceptable and what is not. They read that, then they know how
> they are expected to behave and how they will be treated.

Formal CoC and informal descriptions carry about the same weight in that 
respect: A description of what the group believes about itself.

It's only marginally related to how the group actually acts.

 > They are
> much more likely to be a part of that community. Is it the only
> reason people join a community? No, but unfortunately its becoming a
> big part of the decision (especially for those in the minority).

Collect enough attributes and you're guaranteed to be in some minority.

If it's really a concern, pretend you're a white Western male. It's easy 
enough on a mailing list.
This doesn't work on conferences, but as I said elsewhere: That's 
covered by house rules. Select locations that do not accept harrassment 
and you'll be fine.

> We can argue into we are blue in the face about which is more
> effective in preventing abuse. There is no way to prove this 100% one
> way or another. But there are other goals of a CoC that can not be
> ignored and are not met by the alternative CoC.

On the other hand, the more explicit a CoC, the more it becomes a tool 
for the power-hungry.
Don't laugh, don't shrug that away. Read Orwell's "Animal Farm" instead, 
it's a parable of good intentions being subjected to piecewise 
subversion, until the same privileged positions are re-established, just 
with different people holding them.

At best, the CoC will be ignored and the list will stay the way it is.
At worst, the CoC will be wielded as a tool of power.
The actual content of the CoC doesn't matter much for these things; what 
matters is how people act, not what CoC they claim to follow.

More information about the llvm-dev mailing list