Re: PostgreSQL committer history?

On Wednesday 08 March 2006 18:52, Neil Conway wrote:
> On Wed, 2006-03-08 at 17:52 -0500, Robert Treat wrote:
> > I think Bruce's take is more accurate. For example, look at folks like
> > Dave, Magnus, Teodor, or myself; none of us have commit (afaik) but I
> > would like to think we would all be trusted not to screw things up if we
> > had it.
>
> Teodor does have the commit bit, but only for GiST, tsearch, and related
> code.
>

Woops.. is it Oleg who doesn't have it? ISTR one of those guys didn't. But
more to the point, we don't have granular commit bits as your implying afaik.

> It's not up to me, but to be frank I *wouldn't* trust you, Dave, or
> Teodor with the commit bit for the bulk of the Postgres tree because
> IMHO you haven't modified the tree enough to deserve that degree of
> trust.

That seems goofy IMHO. If the commit bit is to signify "trust", then ISTM
that you would give it to those people whom you trust not to screw something
up. Taking myself for example, if you give me commit I'm not going to start
trying to sneak optimizer changes in. Really there probably isn't any of the
C code that I would change without first sending a patch... but I think I
have modified enough of the doc code and faq's to be comfortable
committing... although even there I might send patches in... heck I still
send in patches for phpPgAdmin some times just because I think it is good
practice. There are a few minor things like typos and what not that I have
seen that I wouldn't waste the time on to send in a patch that I would fix if
I had commit...

> (I'd personally be happy giving Magnus the commit bit, but core
> tend to be conservative about handing it out.)
>

Yeah, sometimes I wonder if giving a few more people commit would be a bonus.
Not so much that we're all feeling constrained that we can't get are favorite
sort methods put in, but for picking up some of the little things; there's a
number of people whom I would trust not to screw things up, and anyway
everything can be reversed and commit can be taken away if you need; we're
all publicly accountable in that regard.

> > OTOH I guess there might be more people like you who look at it like a
> > trust thing, and I just haven't been told about this since I'm not
> > trusted. :-)
>
> Well, on what basis do you think -core hand out the commit bit?
>

Something along the lines of frequency of work on the main trunk? Where it is
more practical for a developer to just have commit than for them to funnel
through core, core hands out the bit.

> > Given the amount of access I have to other things, I doubt that's the
> > case though.
>
> Access to other things is irrelevant: we're talking about the right to
> directly change the source code. There is no reason why the people who
> are trusted to maintain the website ought to be trusted to commit
> unreviewed patches, or vice versa.
>

It all depends on what your trusting. If its a trust in someones technical
ability, then I would agree with you, the two are really orthogonal. That
said I don't recall Marc (sorry Marc) being the flex/bison wizard, so it
really must be more than that eh? That's why I say if it was really about
trust, it should be trusting them to contribute in a meaningful way without
screwing things up. There are a number of people who manage servers who
could cause far more havoc than just getting the commit bit on the source
code, really that's almost minor. I mean there aren't any secrets here...
any change you make to cvs is going to be mailed out to a public mailing list
and archive. If you screw things up, your access is going to get yanked.

--
Robert Treat
Build A Brighter Lamp :: Linux Apache {middleware} PostgreSQL