db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Kathey Marsden <kmarsdende...@sbcglobal.net>
Subject Re: next 10.7.1 steps
Date Tue, 16 Nov 2010 23:52:37 GMT
On 11/12/2010 7:53 AM, Rick Hillegas wrote:

Thank you Rick for your work on the release and Derby in general and 
sharing your thoughts on testing and releases.  I've been thinking about 
what you said and thought I would share a few of my own thoughts below.

> Buddy-testing = Integration test
I  disagree here. I  think buddy testing is just that, another 
developer, looking at a feature and the documentation with a fresh 
perspective  in an ad hoc way.  It used to be that a community member 
would incorporate new features and data types into nstest, mailjdbc, 
sttest, and other tests to provide integration testing.  Sadly, nobody 
seems to have bandwidth for that anymore,  so we don't have the benefit 
of good integration testing, but I don't think that would claim buddy 
testing is a replacement.  We should acknowledge the hole.

> Release candidate vetting = Alpha test
> First release from a new branch = Beta test
I think any release we make needs to be production quality, especially 
with regard to data integrity and upgrade.  If a beta is required, we 
should vote on a beta release, with all the normal disclaimers that go 
with it.

> Second release from a branch = Production-quality release
Everyone has their own comfort level in terms of how long to wait before 
upgrade to a new branch,  but I think if we make a release it should be 
production quality or if we think it is not, we should label it 
appropriately as beta.  Pushing quality standards down the line like 
this can have a cascade effect, where of course whoever used to wait for 
the second release, will now wait for the third,etc.

It was sad I thought when we could no longer release snapshots and 
release candidates to the user community as we lost lots of alpha and 
beta testers and it seems contrary to the whole open source model to me, 
but given that restriction,   perhaps we just need to adapt with a 
lightweight process on alpha and beta releases and label them as such.  
I think the main thing is that we have to have a vote. Never do I think 
we should call something a release that we think is not worthy of the title.


View raw message