Updates-testing (was: Re: thunderbird upgrade - wtf?)

[Alex Hudson]

On 14/10/09 15:31, Jesse Keating wrote:
> On Wed, 2009-10-14 at 09:27 -0500, Mike McGrath wrote:
>
>> The problem isn't GLODA and smart folders, it's that we have no process in
>> place to identify and deal with problems like this before it's too late.
>>
> Aside from updates-testing you mean, where people can test potential
> updates and give feedback as to how they work on their systems?
>

For me, there is a bit of a problem with updates-testing: the machine I 
work on is my primary desktop, and I'm extremely wary of getting myself 
into a situation I can't easily get out of. Now, you might argue that 
avoiding u-t is essentially avoiding the inevitable (and Tbird 3 has 
shown me that, so I agree), but it is riskier.

What would sell me totally on u-t was if there was something where I can 
roll back bad packages.

I'm pretty sure there are various obscure technical reasons why rolling 
back isn't possible in 100% of cases, but I don't think that's 
necessary. So long as it's in the high 90%s then it's good enough, and 
to be honest I would want to avoid testing updates I can't revert like 
the plague anyway: not being able to roll back to my mind is a good 
indicator of not being suitable for a stable release.

In my ideal world, PackageKit would update my stuff with testing 
updates, and anything which broke could be reverted back to some 
previous date or something - either by package if I can identify it, or 
by actual last-known-good date. I'm sure that's a tonne of work, but 
that's the only way I can see the testing system making sense for people 
who rely on their Fedora desktop.

Cheers

Alex.