[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: 1 year release good enough.

On 01/01/12 18:12, Henrique de Moraes Holschuh wrote:
On Sun, 01 Jan 2012, dE . wrote:
I was wondering about the 2 year release cycle of Debian and it's
adaptability on the Desktops.
We cannot do 1 year, it is not enough time to get hard things done
(remember: Debian is _very large_), and still freeze for enough time to
get things right.

1.5 years is probably doable, maybe.

If workload is the issue, then definitely it's not a good idea.

You have to admit that Debian is not used used much on the Desktops
It is the basis of several widely-used desktop distros, and an important
part of Ubuntu.

My main issue was the repository, not the PM.

-- it appears to be more popular for servers; and the 2 year release
cycle is good for servers; increasing the release cycles to a higher
The 2 year release cycle is good for _us_ to get it ready for a stable
release, it is not that large because of servers at all...

On the desktops however, in the above context, things differ completely.
There's new hardware available always; within a period of 2 years,
We refresh hardware support every year (stable-and-a-half), although to
a more limited extent than we do in the testing/unstable distros.

Ok, I didn't know that.

As a consequence, I suggest a sub-stable branch who's release cycle
will be 1 year. As compared to the stable branch, this branch should
Maybe you can morpth this to a deep-stable-freeze-like stabilization of
key parts of the distro taken from testing and backported to -stable
(X.org, kernel, some other stuff) in order to reduce freeze surface, in
order to make the stable-and-a-half releases still just as safe, but
more useful to the desktop...

I think that's a more realistic goal, with better chances of
implementation.   The key word here is to work with a reduced set of
packages that will be allowed large updates and therefore will require
the usual extreme amounts of testing we do before a stable release.

If I understand this correct, there's not much use of this since the backport branch is always there...

Reply to: