Hi Andrzej,
The current process seems to be that we release often and then wait
some time before declaring a particular release as stable. So I see
no problem in releasing something sooner than later that contains
fixes for major defects that we currently know about. It will then
'soak' in the community for some time before we declare it stable or
discover we need a follow-on bug fix release.
Make sense?
Cheers,
Nige
On Mar 13, 2007, at 12:50 PM, Andrzej Bialecki wrote:
Dennis Kubes wrote:
Can we get the new HADOOP-1115 included in this release? This
would really help with allowing Nutch to use 0.12.1 as opposed to
the older 0.11.x release
Why are you (the Hadoop guys) pushing for the release so quickly?
Are you under a deadline? IMHO things should be allowed to settle a
bit - I keep having a distinct impression that the code is unstable
now, and needs to be tested more thoroughly before proclaiming its
release-worthiness - and IMO delaying the release by 24 hours won't
help much.
Having that many critical issues 24 hours before the release
doesn't give me too much confidence in the quality of this release
- I'm more and more convinced that Nutch should use 0.11.2 for now.
I understand the principle of releasing early and often, but there
has to be some point in making a release, i.e. the code should be
mostly usable - as I see it now there is a constant stream of new
critical issues being uncovered with the current code.
May I suggest that you should postpone the Hadoop release by 2
weeks or so?
--
Best regards,
Andrzej Bialecki <><
___. ___ ___ ___ _ _ __________________________________
[__ || __|__/|__||\/| Information Retrieval, Semantic Web
___|||__|| \| || | Embedded Unix, System Integration
http://www.sigram.com Contact: info at sigram dot com
