[Webtest] [dev] roadmap or who is doing what

[Paul King]

Yes, whatever we do for release management shouldn't be too big a
burden on committers or force them to alter their behaviour too
drastically. We could adopt something similar to ant but maybe not
as formal as they have Apache rules to follow. If we did something
like ant but simplified it would involve a process something like:

(1) someone proposes a freeze date (about a weeks notice) and
release date (about 2 weeks after freeze date) and propose
what they think needs to be done to finish off the release
[similar to Mittie's recent email]

(2) we can all +1 or -1 or 0 to indicate if the timing would
impact any planned changes (and/or whether we could assist
with the release) and we can reschedule the date(s) if needed.
We can also debate which things should be in before we release

(3) when freeze date arrives we make a branch/tag for the release
(at that point any changes for the release would need to be merged
onto the branch and HEAD, whereas any changes not destined for
the release would just go to HEAD)
[this addresses Marc's concern as anyone without time can simply
ignore releases and commit to HEAD]

The only question is whether cruise knows about releases. For the two
weeks during the freeze do we get cruise to build "release x.y RC"
releases and anyone wanting HEAD would need to do manual builds
or do we keep cruise on HEAD and make the x.y release manual once
we have all the Jira issues that we promised fixed.

We have lots of options - we should try to keep it light weight. :-)

Paul.

Marc Guillemot wrote:
> I agree to most of your arguments except concerning issue priority: it's 
> an open source project and therefore mostly made by people on an unpaid 
> base (more precisely they are paid to do their job in which they use 
> webtest, but not paid to develop webtest as it). Therefore feature will 
> surely continue to come "because [I] need it". We can and should define 
> a roadmap for "the big picture" but for my part I can't imagine keeping 
> for myself something I've added to webtest because I absolutely needed 
> it even if it's not in the roadmap.
> 
> Marc.
> 
> 
> Paul King wrote:
> 
>> Dierk Koenig wrote:
>>
>>>> I still think more frequent releases might help.
>>>
>>>
>>>
>>>
>>> I agree.
>>>
>>> What timeframe do you have in mind (weeks or so)?
>>>
>>> cheers
>>> Mittie
>>
>>
>>
>> I think something like 3-6 months for each release
>> with the ability to release quicker if we need (e.g.
>> to fix some serious problem or if a large amount
>> of functionality was added at a particular point
>> in time). I am not religious about the exact time
>> but that timeframe feels about right to me for what
>> I think will benefit us and our customers the most.
>>
>> My thinking regarding benefits to customers:
>>
>> Early adopters are always going to go for the latest
>> release and they will be happy to do their planning
>> based on CVS histories and their own debugging.
>> However, the rest of WebTest's customer base
>> probably need to plan introduction of new versions
>> of software and having releases allows all concerned
>> to gain confidence in the functionality that particular
>> releases provide and better plan migration to particular
>> releases.
>>
>> My thinking regarding benefits to us:
>>
>> Having a release gives us the chance to pause from
>> our normal "add a feature because I need it" approach
>> and think through what makes sense to include in a
>> release and force us to try to get our house in order.
>> It also allows us to prioritise issues by releases.
>> We can allocate fixes to particular releases and know
>> what is coming up. It also gives us a greater ability
>> to celebrate our achievements if we choose.
>>
>> Cheers,
>>
>> Paul.
>>