lwn article on the death of Fedora Legacy

[Tim Thome]

Jesse Keating wrote:
> On Thursday 19 October 2006 11:44, Matthew Miller wrote:
>   
>> I think this is really unfortunate, because it makes a big gap in the
>> Fedora ecosystem. This will be largely filled by migration to RHEL-rebuild
>> distros like CentOS, which is well and good (and particularly painless from
>> the end-user point of few) but bad for Fedora.
>>
>> Without a functioning lifespan of over a year, Fedora is only practically
>> useful as an enthusiast, bleeding-edge distro. That's only supposed to be
>> _part_ of its mission.
>>     
I can't speak for others, but going into Fedora Core, I knew what the 
limitations were, and I adjusted my expectations accordingly.

I think what much it is boils down is to what is Fedora?

It's a distro that is close to the tip of development on GNU/Linux, 
close enough to be cutting edge, but not so close to the tip to be 
useless. I knew this going in, and Fedora has done well for what I 
expected it to do. It's fairly stable, has up to date items for most of 
the things that I'm interested in for development, and let's me explore 
some of the items that are next step for RHEL...

The realistic expectation is that SQA has been done for core and 
updates, along with extras... and that Fedora is not forever...

If a longer life-cycle is desired, move over to RHEL/Clones... you'll be 
happier for it.

> Here is what I think can happen.
>
> A) Kill off RHL now.  Stop trying to do stuff there when we just don't have 
> the man power or the volunteers.
>   
Agreed, might get some flak here from others, but is Fedora Legacy the 
right place for supporting RHL?
> B) Move to using Extras infrastructure for building packages.  They're ready 
> for us for FC3 and FC4.
>   
Again, agreed... can prolong things to some extent...

> C) Move to Core style updates process.  Spin a possible update, toss it 
> in -testing.  If nobody says boo after a period of time, release the darn 
> thing.  If somebody finds it to be broken, fix it and resubmit.
>   
For non-critical patches, this is more than fair...

> Somewhere in there convince Luke Macken to do the work to get a Fedora Update 
> tool available for use externally that does the boring stuff like generate 
> the email with the checksums and with the subpackage list and all that boring 
> stuff.  It could even handle moving the bug to 'MODIFIED' when it goes in 
> updates-testing, and finally to CLOSED when it goes to release.  Then it 
> would be easier to get people to contribute, as they'd just be doing things 
> like checking out a package module, copying a patch from somewhere, commit, 
> build.  That would help a lot.  Somebody more "senior" in the project would 
> fiddle with the tool to prepare the update, and do the sign+push.
>
> I honestly think that doing these things is the only way that Legacy will 
> survive.
>   
What would be nice, in a perfect world, is that we change things...

Dev/Stable/Maint... add one more level... Maint would be [security] 
updates only for -2 from current, Stable would be the previous release, 
and Dev would be the current release...

on the eve of FC6 release (hopefully)...

Dev - FC6
Stable - FC5
Maint - FC4
Obsolete - FC3 and earlier...

And then increment once the next Development snapshot is formally 
released...

Keeping that in mind, this reduces the load to Legacy, as Legacy can 
work through the maintainance of non-core/non-security updates, and this 
prolongs the Legacy releases...

My biggest grip right now with moving from one snapshot to the next 
(i.e. from FC3 -> FC4 or FC4-> FC5) is that upgrading is not very clean.

Sorry if this is a bit late... been busy in the real-world, but this is 
something that we can fix... both for supporting older releases as well 
as making the migration less painful...

Tim