Buyer Beware: A Major Change in NFS is about to happen

[Toshio Kuratomi]

On 09/29/2009 05:38 PM, Steve Dickson wrote:
> On 09/29/2009 08:17 PM, John Poelstra wrote:
>>>> [...snip...]
>>>>
>>>> I want to be perfectly clear that I'm not sounding an "all clear" on
>>>> this by any means.  If your answer here means that this change hasn't
>>>> been thoroughly tested, you're going to have a hard time convincing
>>>> anyone that it should be turning over on Beta freeze day.
>>>>
>>> By no means did I interpret that at all... but here lies the
>>> problem... I had no idea I would  have to convenience *anybody*
>>> of *anything* because I thought I made the dead line... again all
>>> following was the schedule in:
>>>         
>>> http://www.linux-archive.org/fedora-development/372823-all-features-need-100-beta-freeze-2009-09-29-a.html
>>>
>>>
>>> And Today I am %100 finished... Please point out which part of that
>>> did I misinterpret, because the last thing I want to do is cause
>>> problems...
>>>
>>> steved.
>>>
>>>
>> Thanks,
>> John
>>
>> https://www.redhat.com/archives/fedora-devel-announce/2009-July/msg00027.html
>>
>>
>> Many of us were assuming that "testable" and "significantly complete"
>> would be enough to imply that a change like this should be done so it
>> could be tested during the Alpha.
>>
>> How should we word things differently in the future so it is clearer?
> Again, John, that is a wonderful question... From my perspective I
> had my head down working as hard as I could to make this deadline... 
> I knew about the alpha deadline and this dealing... 
> 
> I my past, added things of this size in a beta release was actually
> common.. In alpha release you get the software married to the hardware
> (i.e. barely booting) and in beta release you added everything else... 
> Now the post beta release is when the door close... nothing but bug fixes..
> 
One thing I think is unclear this cycle is the usage of the word "Beta".
 It's been said many times that beta is not really beta but actually
final freeze.  For instance: "If all goes as planned the Beta
(previously known as "Final Development") Freeze" in the message steved
linked to.  And yet, no one actually expects that beta is the final
development freeze.  Maybe we should just give up and not use the beta
moniker for it.  As inaccurate as it is, "release candidate" is a
*better inaccuracy* in terms of communicating to developers.  Developers
understand that by "release candidate" we're going to be very tough
about getting changes in, even if they appear to improve the experience.
 Some things will just have to be deferred after a release candidate is
out the door.

The downside of "release candidate", as Jesse has pointed out before, is
that it communicates the wrong thing to end-users.  There's zero chance
that this cut is going to be the absolute final set of bits that we use
on release day so from an end-user perspective, it's not a release
candidate.

One possibility is to have a beta1 at feature freeze and a beta2 now.
This helps developers realize that nothing but bugfixes should be going
in from feature freeze on but lets end user's know that the release
we're making now is still not finished.  It still doesn't give the sense
of urgency and finality that "release candidate" does, though, so I'm
not sure if that's enough.

Another thing that seems to be apparent in steved's complaint is that
what the expectations for Final Feature Freeze (in July) are not clear
enough.  What is meant by testable?  It seems that we mean, the Feature
should be in, enabled, and in final form at that point.

We want to be able to do integration testing from that point forward.
That means we're no longer testing the Feature in isolation, but as part
of the whole experience of installing and running the distro.  Bugfixes
can be applied but nothing that changes the basic shape or architecture
of the distro as a whole.  Plainly, changing the default from nfsv3 to
nfsv4 should happen at Feature Freeze rather than now for that testing
to be performed.  But there's other meanings of testable.  I'm sure that
steved considered the Feature testable at Feature Freeze... he just
wasn't applying the idea that it needed to be testable as part of the
whole distro experience.

Perhaps we need to list some examples of things that should not happen
after Feature Freeze as well as expectation of what should happen.

Examples of what to do and not do from this point forward:
Do: Have something testable
Do: Have the the feature significantly complete
Do: submit bugfixes
Do not: Enable the feature by default
Do not: Make changes that cause other software to have to make changes

et al.

-Toshio

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 197 bytes
Desc: OpenPGP digital signature
URL: <http://listman.redhat.com/archives/fedora-devel-list/attachments/20090929/42214c1b/attachment.sig>