F9 Feature Process

[John Poelstra]

After FESCo discusses and votes on the new development process proposed by Jesse Keating
at http://fedoraproject.org/wiki/ReleaseEngineering/DevelopmentChangesProposal tomorrow I
would start a new discussion culminating in a vote by FESCo one week from now (November 8)
as to how we want the Feature process to work for Fedora 9.

I gave some initial thoughts are here:
http://poelcat.wordpress.com/2007/10/16/fedora-8-feature-retrospective/

Overall I believe we can roll most of the process forward for F9, however a few tweaks and
clarifications to the process will help things run smoother and make my job easier because
then I can help guide the process in a way that is consistent with what everyone wants and
agrees to.

The current policy is here: http://fedoraproject.org/wiki/Features/Policy

To start the discussion, I have listed the specific issues I think need to be addressed. 
Other input and opinions are welcome too!  I think this threads the conversation better
and keeps all the information in one place.

http://fedoraproject.org/wiki/Features/F9PolicyReview.

Thanks,
John


~~~~~~~~~~~~~~ Quick Overview if you don't want to visit the wiki
~~~~~~~~~~~~~~~~~~~~

Here are some of the big issues from my perspectiveones:

1) What is a FEATURE and how should the policy be clarified
http://fedoraproject.org/wiki/Features/Policy#definition
  -considering that anyone can commit new code and packages to Fedora what is unique about
a feature?
  -it is waste of everyone's time to have someone create a feature page and then for
FESCo to say "denied--that isn't a feature"

2) Do we have common agreement on the purpose of creating a feature page?
  -marketing?
  -testing?
  -release notes?
  -other?
  -all of the above?

3) What if someone creates a nice enhancement to Fedora that is new in the upcoming
release but no feature page is created for it?  For example, system-config-firewall in
F8.

To someone not familiar with Fedora wouldn't they expect
http://fedoraproject.org/wiki/Releases/8/FeatureList and
http://fedoraproject.org/wiki/Releases/8/ReleaseSummary to contain the same information or
for there to only be one page? The ReleaseSummary page is really good by the way!

4) How do we (realistically) encourage people to keep their feature pages up to date?  It
was a pain to keep hounding people to keep the status of their pages up to date--very few
people kept them updated every 2 weeks.  Considering how short our release cycle is I
don't think we can go much longer than that.

5) What does "Feature Freeze" really mean? 
  --Do we want to be more disciplined? If we aren't feature frozen are we willing to
move the end of the schedule to allow adequate testing time?
  --What is the point of having a milestone that we don't really follow?

6) What if a feature is not ''complete'' at Feature Freeze?  
  -How do we decide to drop it versus give it more time?

7) How do we define a feature as ''complete''?

[Matthias Clasen]

On Wed, 2007-10-31 at 19:41 -0700, John Poelstra wrote:

>  ~~~~~~~~~~~~~~ Quick Overview if you don't want to visit the wiki
> ~~~~~~~~~~~~~~~~~~~~
>  
>  Here are some of the big issues from my perspectiveones:
>  
>  1) What is a FEATURE and how should the policy be clarified
> http://fedoraproject.org/wiki/Features/Policy#definition
>    -considering that anyone can commit new code and packages to Fedora what is unique
> about a feature?
>    -it is waste of everyone's time to have someone create a feature page and then for
> FESCo to say "denied--that isn't a feature 
I think it is worth making it more explicit that there are different
kinds of features. Something like "replace program a with program b" or
"rewrite program c" (examples are esd -> pulseaudio, or NM) is very
different from "gradually improve fedora wrt. to foo" (examples are
better laptop support, less power consumption). For the former, it makes
sense to have a contingency plan, which will usually be "stay with the
old version". For the latter, it is fine to just do as much as you get
done for this release, and continue improving the situation afterwards.
For the gradual improvement type features, the main question at feature
freeze time is if we have achieved enough of an improvement to make a
big splash about it.


>  2) Do we have common agreement on the purpose of creating a feature
> page?
>    -marketing?
>    -testing?
>    -release notes?
>    -other
>    -all of the above? 
Pretty much all of the above, I'd say, plus some more
    - developer synchronization if several people work on a feature
    - cross-distro and upstream communication; I find it e.g. a big 
      help to look at Ubuntu blueprints occasionally to see how their
      plans line up with ours


>  To someone not familiar with Fedora wouldn't they expect
> http://fedoraproject.org/wiki/Releases/8/FeatureList
>  and http://fedoraproject.org/wiki/Releases/8/ReleaseSummary to contain the same
> information or for there to only 
>  be one page? The ReleaseSummary page is really good by the way! 
I would expect FeatureList to be the starting point, and ReleaseSummary
the end result of turning it into a shiny and attractive document.


>  6) What if a feature is not ''complete'' at Feature
> Freeze?  
>    -How do we decide to drop it versus give it more time? 
This really depends on the type of feature; for the gradual features discussed about,
dropping them is not very 
meaningful. For "replace existing functionality" or "new stuff"
features, dropping could be a possibility, but 
looking at this at feature freeze time is really too late; if we want to be serious about
dropping features, we 
need to do closer monitoring of features before the freeze, and ensure that the
contingency plans are in place
and that people are aware of the approaching freeze, and don't miss it by accident. 

Out of interest, I didn't see any information about how we did in terms of features
that made it in vs features 
that were dropped or deferred in your F8 recap. Did we drop anything, and if so, have you
looked at why ?


Matthias

[John Poelstra]

Matthias Clasen wrote:
>  Out of interest, I didn't see any information about how we did in
> terms of features that made it in vs features 
>  that were dropped or deferred in your F8 recap. Did we drop anything, and if so, have you
> looked at why ?
>   
Thanks for all of your feedback.  I will make sure FESCo considers it 
when they discuss changes to the policy.

In terms of stats on features that were dropped or deferred--I did not 
actively track them, but that is in interesting question.  Would there 
be value in doing that or is anecdotal evidence enough?  I ask because 
tracking it is all manual :)

In almost all of the cases, FESCo and the feature owner(s) agreed that a 
particular feature wasn't ready enough and should be deferred a the time 
approval was being voted on OR features with stale statuses that I 
raised at FESCo after not hearing back from the feature owner when it 
was pretty much known that that particular feature was not going to be 
completed.

Then there were a couple of features where the feature was approved, 
nothing happened with it and then it was dropped because the owner said 
it wasn't really considered a feature... 
http://fedoraproject.org/wiki/Releases/8/Bookmarks

John

[Matthias Clasen]

On Mon, 2007-11-05 at 16:37 -0800, John Poelstra wrote:

>  
>  Then there were a couple of features where the feature was approved, 
>  nothing happened with it and then it was dropped because the owner said 
>  it wasn't really considered a feature...  
I hope that the nicely done F8 release summary will really drive home
the value of doing the feature tracking and more people will come
forward to do some advertising for the stuff they are doing.
Just comparing the numbers, the recently published feature list for the
next Ubuntu release was some 130 items long, while I currently see ~10
items in CategoryFeatureFedora9. Admittedly, there is probably some 75%
or so in the Ubuntu list that is just hot air or nice ideas that will
never get done, but still...

Matthias

[Jeremy Katz]

On Mon, 2007-11-05 at 19:45 -0500, Matthias Clasen wrote:
>  On Mon, 2007-11-05 at 16:37 -0800, John Poelstra wrote:
>  > Then there were a couple of features where the feature was approved, 
>  > nothing happened with it and then it was dropped because the owner said 
>  > it wasn't really considered a feature... 
>  
>  I hope that the nicely done F8 release summary will really drive home
>  the value of doing the feature tracking and more people will come
>  forward to do some advertising for the stuff they are doing.
>  Just comparing the numbers, the recently published feature list for the
>  next Ubuntu release was some 130 items long, while I currently see ~10
>  items in CategoryFeatureFedora9. Admittedly, there is probably some 75%
>  or so in the Ubuntu list that is just hot air or nice ideas that will
>  never get done, but still... 
I also suspect that our list is just gearing up as we only _just_
finished the release as of late Friday :-)  And there's another ~ 25-30
in CategoryProposedFeature.  I think we'll end up with a pretty healthy
list for Fedora 9.

Jeremy

[Jon Masters]

On Mon, 2007-11-05 at 19:45 -0500, Matthias Clasen wrote:

>  I hope that the nicely done F8 release summary will really drive
> home
>  the value of doing the feature tracking and more people will come
>  forward to do some advertising for the stuff they are doing.
>  Just comparing the numbers, the recently published feature list for the
>  next Ubuntu release was some 130 items long, while I currently see ~10
>  items in CategoryFeatureFedora9. Admittedly, there is probably some 75%
>  or so in the Ubuntu list that is just hot air or nice ideas that will
>  never get done, but still... 
* 131). Send all the developers into space at least once.

:-)

Jon.

[Rahul Sundaram]

John Poelstra wrote:

>  To someone not familiar with Fedora wouldn't they expect 
>  http://fedoraproject.org/wiki/Releases/8/FeatureList and 
>  http://fedoraproject.org/wiki/Releases/8/ReleaseSummary to contain the 
>  same information or for there to only be one page? The ReleaseSummary 
>  page is really good by the way! 
Feature list is more developer oriented. Release Summary is a subset of 
the feature list highlighting major features and "themes". If there are 
for example 3 user interface related features, 4 security related ones 
and 2 performance related ones, we want to highlight them together in 
the release summary.  The actual release announcement is sort of like a 
easter language and is not easily translatable and we needed an 
alternative that was closer to a press release. It is in fact the press 
release writing attempt that was combined with the release notes 
overview and morphed into the release summary for Fedora Core 5 and 
since it turned out be fairly successful (a lot of users and reviews 
referenced that), we continue doing it.

Rahul

[Christopher Aillon]

John Poelstra wrote:
>  4) How do we (realistically) encourage people to keep their feature 
>  pages up to date?  It was a pain to keep hounding people to keep the 
>  status of their pages up to date--very few people kept them updated 
>  every 2 weeks.  Considering how short our release cycle is I don't think 
>  we can go much longer than that. 
Before we ask this question, first ask what do we gain vs. lose from 
people updating the page every two weeks?  Do we gain anything?  Not 
much, really.  We get a false sense of security that it will get done on 
time or it won't get done on time.  A feature can get lots of work early 
on and then the maintainer(s) go AWOL and leave it at 70%.  Or there can 
be no updates for a long while as they are doing other things and then 
it gets done within the final two weeks before freeze.

And we lose engineering time from having to update the page.  For some 
of our contributors who are not paid to work on Fedora, this could be a 
non-insignificant portion of the time they can allot to Fedora.

Is it really worth getting people to update their feature page every two 
weeks?  I'd argue it's much better to start looking for updates in the 
week or two prior to the freeze as those will be more meaningful.  We 
should also encourage people to update their page regularly (or when 
their feature is completed), but I'm not convinced that enforcing this 
is a good idea.

[Jesse Keating]

On Wed, 07 Nov 2007 15:38:14 +0100
Christopher Aillon <caillon(a)redhat.com&gt; wrote:

>  Before we ask this question, first ask what do we gain vs. lose from
>
>  people updating the page every two weeks?  Do we gain anything?  Not 
>  much, really.  We get a false sense of security that it will get done
>  on time or it won't get done on time.  A feature can get lots of work
>  early on and then the maintainer(s) go AWOL and leave it at 70%.  Or
>  there can be no updates for a long while as they are doing other
>  things and then it gets done within the final two weeks before freeze.
>  
>  And we lose engineering time from having to update the page.  For
>  some of our contributors who are not paid to work on Fedora, this
>  could be a non-insignificant portion of the time they can allot to
>  Fedora.
>  
>  Is it really worth getting people to update their feature page every
>  two weeks?  I'd argue it's much better to start looking for updates
>  in the week or two prior to the freeze as those will be more
>  meaningful.  We should also encourage people to update their page
>  regularly (or when their feature is completed), but I'm not convinced
>  that enforcing this is a good idea. 
I agree that "every two weeks" seems arbitrary.  However we could time
getting updates a week or so before the Alpha, before the Beta, and
obviously more updates as we get nearer the final freeze.

Also it should be made known that we can do special snapshot live and
dvd spins of Rawhide at any point when it would make sense for a
Feature.  If a Feature has hit a significant milestone but it is still
too far away from either Alpha/Beta/rc's or one of the weekly snapshot
attempts after Beta, we can create a snapshot and highlight the Feature
in question.

-- 
Jesse Keating
Fedora -- All my bits are free, are yours?

[David Nielsen]

ons, 07 11 2007 kl. 15:38 +0100, skrev Christopher Aillon:

>  
>  Is it really worth getting people to update their feature page every two 
>  weeks?  I'd argue it's much better to start looking for updates in the 
>  week or two prior to the freeze as those will be more meaningful.  We 
>  should also encourage people to update their page regularly (or when 
>  their feature is completed), but I'm not convinced that enforcing this 
>  is a good idea. 
Having reasonably updatyed feature pages really aids in advocacy. It
gives us a place to point people to as a means of showing what is
coming. Additionally not having the updated before the last few weeks
makes it look like we never do anything, which we can agree is untrue.

Meaning is not just having a finished product, it's marketing it,
getting people interested in the product and it's placing it in a larger
software ecosystem. Please feel free to tell us empty but expensive
sounding words of love when explaining to us how your feature will
improve our lives. We like that.. it makes us all hot and bothered about
Fedora.

I'd say every 2-4 weeks we ping the feature owners who haven't updated
and ask them to take the time to put down a few words on the progress.

Preview writers will have places to look when getting excited about the
next Fedora. Also having clear progress on active features makes them
visible so that when we release, the feature will be known as born and
raised in Fedora.. hopefully a growing commitment to telling the world
on going what we are doing will reducing the amount of reviews I have to
read wherein Ubuntu is credited as the genesis for half our features.
The horrible journalistic skills involved here just makes the situation
even sadder - nobody actually checks their facts and never have I seen a
review get amended when I send in factual corrections. I want our
features praised when they come out.. in Fedora.. I want our developers
to get the attribution they sorely deserve. Please help make that
happen, update your feature page.

- David

[Christopher Aillon]

David Nielsen wrote:
>  ons, 07 11 2007 kl. 15:38 +0100, skrev Christopher Aillon:
>  
> > 
> > Is it really worth getting people to update their feature page every two 
> > weeks?  I'd argue it's much better to start looking for updates in the 
> > week or two prior to the freeze as those will be more meaningful.  We 
> > should also encourage people to update their page regularly (or when 
> > their feature is completed), but I'm not convinced that enforcing this 
> > is a good idea.
>  
>  Having reasonably updatyed feature pages really aids in advocacy. It
>  gives us a place to point people to as a means of showing what is
>  coming. Additionally not having the updated before the last few weeks
>  makes it look like we never do anything, which we can agree is untrue.
>  
>  Meaning is not just having a finished product, it's marketing it,
>  getting people interested in the product and it's placing it in a larger
>  software ecosystem. Please feel free to tell us empty but expensive
>  sounding words of love when explaining to us how your feature will
>  improve our lives. We like that.. it makes us all hot and bothered about
>  Fedora. 
So let's have beat reporters.  Come to the weekly SIG meetings.  Ask 
questions about features.  Write a blog report about the progress. 
Maybe have it in FWN.  Having people write about it in this fashion is 
much more excitement generating than a boring static page that people 
have to manually check every so often.

Engineers want to work on the features not the paragraphs, or maybe they 
aren't good at writing things in an exciting way.  If we really want to 
generate excitement, and get people interested in things, then features 
pages is IMO the wrong approach.