F11 Proposal: Stabilization

[Les Mikesell]

Jeff Spaleta wrote:
> On Tue, Nov 18, 2008 at 2:16 PM, Les Mikesell <lesmikesell at gmail.com> wrote:
>> I think what people really want is 'updates that fix the things that are
>> already broken' but not 'updates that break something new'.  Can you come up
>> with a way to 'crowdsource' this statistic?  Perhaps an optional poll where
>> people could rate the health of their system and individual apps, preferably
>> tied somehow to the smolt hardware reports so someone could see how the
>> current updates run on hardware like their own or which update triggered a
>> flurry of problems.  Or maybe this could be automated - but the absence of
>> problem reports for an update could mean that no machines survived to send
>> them...
> 
> Uhm.... in order to get "crowdsource" stats.. which will help you
> prevent the releasing of updates to the crowd...there has to be a
> crowd of people using the updates. How do yuou get feedback from the
> crowd without exposing the crowd to the updates?

Every time anyone mentions slowing down the feature changes in favor of 
fixing the brokenness there are a flurry of postings from people saying 
they want all the new features they can get.  There has to be a way to 
take advantage of these willing guinea pigs (or is it canaries in a mine 
shaft?) and let their experience determine when it is safe for everyone 
else to follow.  But, you either need an intermediate repository with 
things moved on to the safer one at some point, or a client-driven 
mechanism that knows how to request the degree of vetting desired along 
with some way to accumulate the statistics.  I'd personally be much more 
interested in keeping an up to date test box running if I knew that 
experiencing problems on it would have any eventual benefits.

-- 
    Les Mikesell
     lesmikesell at gmail.com