Re: [debian-devel] Re: equinoctial releases
A levelezőm azt hiszi, hogy Andrew Suffield a következőeket írta:
> On Wed, Sep 01, 2004 at 01:38:21PM -0600, Thomas E. Vaughan wrote:
> > Nevertheless, it seems to me that debian should release
> > twice a year.
>
> No way. That would be *horrible* for all the serious users. We would
> have to support the previous *two* releases, because upgrade ==
> downtime and a large corporate environment won't tolerate upgrades
> more often than once a year, and would rather have them less frequent
> than that.
It is a point. However such environments are a minority of large
corporate ones. They mostly have upgrade procedures tailored to
the properties of the products they use, and some have none.
>
> Supporting two stable releases at once is not feasible. We have enough
> trouble with one.
Some of this trouble comes from the fact that two releases are too
far from each other.
Also I don't think that the majority of developers fix bugs in the
stable release only the really serious ones. The release procedure
and the rules discourage it, even in the cases where it is possible.
>
> 12-18 months is a decent timeframe. Any less is too fast. Anything up
> to 24 months is still comfortable. Longer and it starts to get
> annoying.
When you plan upgrade timeframes, you should also keep the needs of
development in mind. Because the complexity of dependencies fast upgrade
cycles make life easier, and also more encouraging for the developer.
Also a lot of our "potential customers" say that we are too outdated.
I think we need a new approach to releases. Maybe Debian should
be separated to more "products" like OS core, gnome desktop,
KDE desktop, etc, and these products should have their own release
cycles or even approaches to release, like introducing service packs,
continous upgrades, etc.
I guess/hope that an analysis of dependencies would show how Debian
can be separated to such products.
--
GNU GPL: csak tiszta forrásból
Reply to: