BDFL and the review process

[Ian Clatworthy]

Robert Collins wrote:
> On Mon, 2007-07-16 at 11:45 +1000, Martin Pool wrote:

> In terms of it being reviewed here, our review process has a hole in it
> in some regards - and I think that hole is fine, but it does mean that
> sometimes people will be unhappy with what goes in. The hole is that we
> require two +1's, and no -1's, but theres no minimum window on things
> being done, which means that there is potentially a very limited time
> window to say '-1' or even '-0' (which is what I am on this - I don't
> think its sooo ugly as to do -1),

I don't think we can avoid the "sometimes people are unhappy" issue. As
the size of the core group grows, we will increasingly get situations
where one or more people disagree. The world is a lot more grey than
black and white, even though most of us - me included - have black and
white views on most topics.

To take another example, John has voted +1 (conditional) on Kent's
changes to removing the whitespace on the left margin when only merge
revisions are shown. The conditional is not on more code changes - it's
on getting *agreement* that the changes is a good thing. Aaron doesn't
want it but has abstained from reviewing the code again. Stalemate?

Democracies are great but occasionally projects need a benevolent
dictator for life. As part of the changes Aaron is making to BB to
improve our review process (our biggest process issue), perhaps we need
a 'BDFL Intray' for things that Martin needs to hear the arguments on
and adjudicate? We have a *huge* amount to do and sometimes we simply
need to make a decision, accept it and move on. (I have firm opinions so
I know that's not easy BTW.)

I actually think the process is working correctly here. A change was put
up, reviewed and got through. Robert missed it so is correctly
questioning the decision after the fact. That's a better process than
waiting for everyone to sign off and blocking on total consensus.

For the record, it was *me* who probably triggered Martin's patch. I
went into my bzr.dev directory and typed 'bzr selftest' instead of
'./bzr selftest'. With the current output after this change, I would
have realised my mistake. The problem was running the wrong version so
running 'bzr version' only occurred to me *after* the problem was found.
It was a tired mistake, but then again, most are.

Ian C.