Cees de Groot <cg at cdegroot.com> wrote:
> On Tue, 2003-05-06 at 13:48, Andreas Raab wrote:
> > Finally (because I'm just at some larger issues) consider dropping the
> > "micro management" approach for larger projects. This just won't work.=20
> 
> Hear, hear. We're in alpha mode. We're spending more time on discussing
> stuff to include than writing stuff to include. 
...
> I vote to give Diego and Stephane write access to the update stream for
> their projects for the coming two months or so. By then, we will know
> whether it's a good idea or not and we can decide to revoke the access
> for the process of moving to stabler code or not.

IIUC, you advocate giving different people write access to a complex
piece of code, without spending some time clarifying intentions first. I
find it hard to imagine a faster way to degrade what integrity a piece
of code has, using top notch programming talent, all full of good
intentions. 

The day we let people harvest their own code, or let them harvest
without making a serious effort to make sure we're all on the same page
first, I quit. Not what I signed up for.

Cees de Groot <cg at cdegroot.com> wrote:
> On Tue, 2003-05-06 at 16:02, Daniel Vainsencher wrote:
> > No, I think some reviews of the code itself are essential for clearing
> > up misunderstandings. Without a shared history, words are too imprecise.
> >=20
> Absolutely. But why does it have to be one set of 'privileged' eyes?
> Just throw it in the update stream as long as you're alpha, so anyone
> can look and feed back information/comments to the authors.=20
The cost of finding and fixing the mistakes is too high, and will get
higher with time (no reason for people to limit their own changes).
Reviewing first, and giving more freedom later is a pain at first, but
gets lighter later.

> We probably want different rules for different times: alpha accepts
> mostly anything (or maybe mostly anything after the first patch by an
> author), beta needs X amount of review, gamma needs Y amount of review,
> and released needs Z amount of review, where presumable X < Y < Z.=20
See answer to your suggestion above. Sounds like a recipe for disaster
to me, and I know of no project that desire life and is run this way
(cvs access without reviews after any single patch accepted). If you
want to see the other extreme of review policy, see Mozilla's.

Daniel