Re: Is Linux Unix?
John Summerfield <debian@ComputerDatasafe.com.au> writes:
> Brian Nelson wrote:
>
>>Carl Fink <carl@fink.to> writes:
>>>No one seems to like my proposal: six months after any stable release,
>>>freeze, then test and release. Period.
>>
>>It's not that no one likes the proposal. It's more of the "herding
>>cats" phenomenon. There's no consensus on how the release process
>>should work, it's hard to make a thousand scattered volunteers listen
>>anyway, and no one other than the current release manager is willing to
>>step up and actually try to manage the release...
>
> I was thinking about that, thinking back to Mr Worthington, my history
> teacher in the early 60s. Teaching me history then was a pretty
> fruitless task, but I do recall that democracies don't provide efficient
> governance.
>
> That's why we don't have democratic armies. While a democratic army is
> planning the next battle, it would get cut to ribbons by the enemy.
>
> Debian is worse than a democracy, it's a democracy of hackers. No
> managers at all, to speak of.
>
> Hackers are good at cutting code. They don't like documentation - go
> look for debian-installer documentation for example.
> They don't like deadlines. "We'll release it when we're good and ready."
> Hackers who show management talent are likely to go get lucrative jobs
> managing. Managing the hosts of hackers here must be pretty thankless.
>
> The ideal government, according to Mr Worthington, is a benevolent
> dictatorship. A benevolent dictator's word is law, but he rules to the
> benefit of those whom he rules. Taking advice is, of course, allowed and
> good. So is making decisions.
Debian's release manager *is* essentially a benevolent dictator, or at
least he very nearly has the power to be. He takes more of a hands-off
approach and tries to automate the process as much as possible
(automatic migration of packages from unstable to testing). Automation
is a very good thing since it provides excellent scalability.
He doesn't set goals or time frames very often, probably because it's
not very helpful. For example, it took at least 6 months just to get
glibc in releasable shape across all architectures, another few months
for the gcc 2.95 => 3.x transition, and a couple years to get
debian-installer usable. If the RM were to bark out orders and demand
that everything be ready in 6 months, it wouldn't really help anyway
since it doesn't make fixing stuff any easier.
debian-installer has been the major bottleneck, and everyone has known
it would be since woody's release. It doesn't matter how many times the
RM shouts, "Hey, we need to finish the installer!" Someone still needs
to do the work.
> That, I think, is about how the kernel works.
The kernel release process is very similar to Debian's, and they
generally take about the same amount of time (2-3 years) to prepare a
release.
The major difference is that the stable kernel is waaaaay more lax when
in comes to adding new features, especially for hardware support, so no
one really minds when it take 3 years to release a new kernel. Debian
stable, on the other hand, is nearly unchanged since its release over 2
years ago.
--
You win again, gravity!
Reply to: