geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sachin Patel <sppat...@gmail.com>
Subject Re: Frustrations of a Release Manager
Date Fri, 09 Jun 2006 18:31:55 GMT
I just saw this thread and want to say my 2c, I haven't yet read the  
other threads and have to run out so sorry if this statement has been  
repeated.  The most important thing we can do to make the project  
succeed is to ship, and to ship often.  Moving forward we need to  
have a fixed interval of when we release and based on those intervals  
each of us need to be accountable on what we can commit.  We  
desperately need to be able to "give-up" function, meaning we must be  
willing to modify our roadmap of contained items and defer items if  
they cannot be contained within the schedule, rather then push out  
the schedule which seems to be the easy answer.  If we can prove to  
our users that they can constantly expect releases at consistent  
intervals this would be a huge win.

Take a look at the Callisto effort in Eclipse.  Not only has the  
Eclipse project not missed a release date in who knows how long (if  
ever), but now they are releasing 10 top level projects  
simultaneously. I say we learn from the eclipse model, follow it, and  
tweak the process to our needs.

-sachin

On Jun 9, 2006, at 9:34 AM, Aaron Mulder wrote:

> In the spirit of greater openness and communication, please elaborate
> on 'thing have been "quietly" injected into Geronimo'.
>
> As far as I can tell, the main source of the 1.1 delay was that the
> module ID changes (new syntax, groupless or versionless dependencies,
> etc.) caused a ton of problems, in the TCK, the deployment tools, the
> console, and so on.  When the original deadline came, the product was
> not stable enough to ship.  I'm sure that some of the features I've
> worked on have contributed -- mainly the keystore changes, which
> caused some TCK failures until we updated the keystore configuration
> for it.  Still, we've talked about some of the reasons for this, and I
> think we all want to try to make the 1.2 changes more incremental and
> keep the TCK passing at all times to avoid major disconnects as we
> move forward.
>
> As far as the release schedule goes, I'm disappointed that we missed
> the deadline, and then didn't really update our road map...  If there
> was a new target date or plan it seemed pretty informal -- there
> didn't seem to be anything posted to the dev list or the web site, etc
> (though based on Jeff's comments it sounds like there was and I missed
> it?).  Now we're trying to put out a release when our only
> preview/release candidate has been available for less than a week.  I
> contrast that to the SuSE process where there were at least 12
> well-defined test builds (9 or more beta builds and 3 or more RC
> builds) at well-defined interrvals.  As a user, I certainly
> appreciated that I could get and try the latest, submit bug reports,
> check the release calendar for the date of the next test build, get it
> and test the fixes, etc.  I don't think that one build and 72 hours is
> sufficient to convince me that 1.1 is a stable release.  I don't feel
> strongly enough to override a majority opinion, if there is one, but
> I'd like to try a much more SuSE-like release strategy for 1.2 and see
> how it goes.  If that doesn't work so well either, we'll regroup and
> try something different for the release after.
>
> Thanks,
>    Aaron
>
> On 6/9/06, Jeff Genender <jgenender@apache.org> wrote:
>>
>>
>> Bruce Snyder wrote:
>> > On 6/8/06, Aaron Mulder <ammulder@alumni.princeton.edu> wrote:
>> >
>> >> I think it will help to have the schedule of the release.  No  
>> one can
>> >> claim IBM has a secret agenda if the time line is laid out  
>> there.  And
>> >> it's easy to wink if no one has any idea what the deadlines we're
>> >> working toward are.
>> >
>> > I agree with Aaron here - publicity of not only the timeline  
>> (i.e., a
>> > calendar of release schedules maybe) but also the Road Map may  
>> help on
>> > all fronts. IMO we should consider publishing and continually
>> > revisiting both of these items. I know that this won't be a popular
>> > suggestion on the committer side of things because we are a  
>> volunteer
>> > organization, but it would most certainly help our user community
>> > immensely.
>>
>> I have to disagree here.  Although I absolutely agree a roadmap is
>> helpful and trackable, the timeline and release issues that Matt has
>> talked about is clearly an issue.  On these lists, Matt has made  
>> things
>> extremely clear regarding when our releases should be, along with  
>> group
>> consensus, and thing have been "quietly" injected into Geronimo.  I
>> share Matt's feelings and frustrations.
>>
>> Minimally, if we cannot hold to a simple date based on agreement on
>> these lists, a roadmap, although helpful, will surely not be a  
>> panacea.
>>
>> It is also my hope that there are not private emails going around
>> talking about "secret" agendas.  This would dismay me as I fully  
>> expect
>> that we are all adult enough to share our feelings with each other in
>> these lists.  If an email like this is being passed around, then we
>> clearly need to be working on our communication skills and have a  
>> long
>> way to go on learning to work with each other as a team.  I think
>> communication is the primary thing we need to deal with.
>>
>> Jeff
>>
>> >
>> > A wiki page of the Road Map along with a rough timeline would be a
>> > good start. I also think that tying the Road Map to a timeline will
>> > cause people to more closely examine the time a particular feature
>> > might require. But like the Linux kernel release schedule,  
>> determining
>> > any kind of regular release schedule may prove to be quite  
>> difficult.
>> > But IMO it can't hurt to have goals.
>> >
>> > Just my $0.02.
>> >
>> > Bruce
>>


-sachin



Mime
View raw message