Where are we going? (Not a rant)

[Andrew Price]

On 07/12/12 17:48, Matthew Miller wrote:
> On Fri, Dec 07, 2012 at 03:11:31PM +0000, Andrew Price wrote:
>> Ah the ol' Fedora stability improvement thread. It must be Friday.
>> Ok, I'll bite.
>>
>> This sort of conversation often comes and goes without much being
>> done. Usually it consists of debates between three camps:
>>
>> 1. Those who see Fedora as an intrinsically unstable distro which
>> showcases and attracts testing for the latest upstream work
>> 2. Those who want Fedora to be stable enough to become a realistic
>> alternative to Windows and Ubuntu for the general masses
>> 3. Those who want Fedora to be stable enough and supported for long
>> enough to be used as a server OS
>
>
> Hmmm. I don't think I'm in any of those camps. I want Fedora to thrive and
> be used, not as an "alternative" but on its own merits.

Each OS is an alternative to the others. "Alternative" here shouldn't 
imply anything negative, simply that users could happily switch from one 
to the other.

> That includes being
> a general-purpose OS, both on desktops, on traditional servers, and in the
> cloud.

Well that's my hope, too. The items on your list are compatible with 
each other but the problem arises when you add the requirement for 
Fedora to additionally be an unstable mixing pot of the latest upstream 
packages. And that's really what rawhide should be, but since rawhide 
requires so much effort to install and keep running, that's what Fedora 
proper has become.

> While I *do* think we would benefit from a slightly longer cycle (with an
> "security updates only" phase), I don't think that's the only way to tackle
> the problem.

I'm not against extending the cycle in essence. I just don't think the 
*first* step should be to extend cycles and/or support. I think we need 
the ability to tackle as many stability problems as we can, pre-release, 
before we can start thinking about extending life cycles, and that means 
getting people using rawhide day-to-day again. The more bugs we allow 
into our releases by neglecting to test rawhide, the more development 
time we have to spend/waste on fixing packages in the three supported 
releases.

I would link to http://lwn.net/Articles/506831/ but you were quoted in 
that article so I'm guessing you've already seen it :)

> For example, making it so key applications and development stacks could
> easily float from one base OS to the next would make it less of an issue
> when the base OS needs to be upgraded.

Not sure I catch your drift here, but it sounds like it could cause API 
mismatch headaches.

> Making upgrades incredibly painless is another good but different approach,
> making us closer to a rolling release. (I think we're headed that way with
> 'fedup', but it's going to be a little longer for that to shake out.)

Yep, I upgraded my netbook to f18 beta with fedup yesterday without too 
much hassle. Looks promising.

Andy