hadoop-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Tom White <tom.e.wh...@gmail.com>
Subject Re: [DISCUSS] Proposed bylaws for Hadoop
Date Thu, 21 Oct 2010 17:10:27 GMT
On Wed, Oct 20, 2010 at 3:15 PM, Chris Douglas <cdouglas@apache.org> wrote:
> On Wed, Oct 20, 2010 at 10:14 AM, Nigel Daley <ndaley@mac.com> wrote:
>> FWIW, "PMC does not generally operate by majority but by consensus." was given as
a rationale when explaining to an existing PMC member why it was ok to approve so many new
PMC members from a single company.
> FWIW, working at the same company doesn't imbue contributors with a
> shared agenda, neither does it rob them of their ability to
> compromise, relieve their concern for the project's health, nor does
> it diminish their respect for other contributors. If individuals from
> that company display any of those traits, then the PMC should address
> them. Equating association and collusion a priori is inadmissible. And
> insulting.
> On Wed, Oct 20, 2010 at 2:12 PM, Tom White <tom.e.white@gmail.com> wrote:
>> I suggest we make this stronger.
>> By way of comparison, the recently enacted bylaws for Pig
>> (http://pig.apache.org/bylaws.html) have consensus, for example.
> Consensus is equivalent to making the PMC a permanent appointment.
> Discussions would be more civil and more likely to offer compromise-
> we would actually work toward consensus- if intransigence and
> hostility had consequences. Removing people is a last resort. Moving
> the barrier closer doesn't make it more likely, but it aligns
> incentives so rational people will reign in their more intemperate
> comments and, instead, find ways to make progress.
> Many projects require supermajorities or even 3/4 majorities. It's
> sufficiently damning if more than half the PMC thinks an individual is
> so destructive that less damage would be done by ejecting them, in my
> opinion, but again: this is not a facility we should use, or have to
> use. Throwing someone out is an expensive failure for the project, and
> an conspicuous failure of its governance and community. Personally, I
> don't care if it's a majority or a supermajority, but consensus is a
> fig leaf.

Fair point. For a large PMC, that's probably true. This matter was
discussed on the Pig list, and there was a range of opinion there:
Personally, I think a supermajority strikes the right balance.


>> Also, Pig has a minimum length of time for each action, rather than
>> the same fixed number for each. Might be worth considering.
> IIRC, the reasoning for a full week in the last thread was to prevent
> national holidays from affecting the definition of "working days." We
> might expedite votes for security/patch releases. -C

View raw message