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

Renato Golin via llvm-dev llvm-dev at lists.llvm.org
Fri May 6 15:13:46 PDT 2016

On 6 May 2016 at 22:57, Tanya Lattner <tanyalattner at llvm.org> wrote:
> The major weapon of harassers is arguing whether something is actually
> harassing. It is difficult to enforce a CoC if you have to have a month long
> nasty argument about whether it was violated. It burns out people like you.

The major weapon about enforcers is *not* wanting to argue.

Harassment is a very complicated issue that involves not only actions
but points of view.

For instance, you thought Joachim was joking, I didn't. If he was
joking, that would have been very bad taste. If he was not, that would
have been a perfectly valid and non aggressive way to understanding
the range of your statement.

Depending on your *point of view*, from that point forwards, things
would roll down in completely opposite paths.

If you don't discuss what is and what isn't harassment, or if you're
not prepared to understand that communications will breakdown very
easily, then you're taking an extremist point of view in regards to
the code and giving every one that has had fears so far, the right to
do so. I'm not saying *you* are, but that is the consequence if you
(or anyone else) would have been.

Now, assuming that he was joking, that would have been bad, but how
bad? It would depend on how many times it happened in the past (serial
offender, see the autism page I shared earlier), or if the affected
person asked him to stop and he didn't, or if he did this on the list
or out of it, etc. Those are all parts of the spectrum of harassment,
and they have to be dealt with in *very* different ways.

If we advocate on our CoC that we do not take a spectrum approach to
harassment, and that some unnamed people are the ones that choose what
is harassment and what is not, than you're telling all prospective
members that we essentially don't care about their opinions and their
behaviours will be monitored and potentially curbed at our own

One may think it's ok to do that to potential offenders, but you
cannot tell who is an offender and who is not just by looking at which
T-Shirt they're wearing. Doing so is the most classic form of

I personally do not see this as the role of the CoC, and from my
conversations with Chandler and others on this list and the review,
this is not what they are thinking either. I'd rather deal with
offenders *after* they have offended than risking have my prejudice
affect the quality of the tools I build. It would also show how
efficient our process is against harassment.


More information about the llvm-dev mailing list