I just saw the re-tagging to 1.0.0, as if we know there are problems
and are setting the stage for 1.0.1. (There's been no public
discussion of this as far as I can tell...)
I've never seen anything like this before, doing a "1.0.0" with
knowledge aforethought about 1.0.1. If we know there are problems,
why not get the fixes in there? I'm not talking about perfection, or
every feature on every wish list, but getting confidence over even a
week of testing would be worth a lot.
Dain was adamant about it, but has so far not explained why it was so
important to rush this out like this. I appreciate the stellar
amount of work going into this, and I'm aware that I'm not doing it,
but still - is anyone willing to answer why it needs to be pushed out
on this timetable?
People need to start trusting software that we release, and IMO
putting out a 1.0 that we know isn't ready doesn't get us closer to
that goal of community trust. There's been a little under 2.5 years
of tremendous passion and energy from many many people that went into
this.
geir
On Dec 18, 2005, at 6:39 PM, Aaron Mulder wrote:
On 12/18/05, Geir Magnusson Jr. <[EMAIL PROTECTED]> wrote:
release as what? rc1 or 1.0?
That was not specifically discussed, but my sense is 1.0 because now
that I look, Matt referred to putting other fixes in 1.0.1.
Thanks,
Aaron
On Dec 18, 2005, at 6:16 PM, Aaron Mulder wrote:
Alan, Matt, and I spoke about the release on IRC. Matt's thoughts
are:
1. integrate the shell script changes to reduce verbosity. Matt
thinks
this is important because its a customer's first impression and an
ECHO OFF is fairly trivial (famous last words)
2. Integrate the fixes from JGenender for clustering since this is
such a big thing for users and this was advertised to folks. The
change looks reasoinalbe...doesn't seem to risk TCK testing and
if it
doesn't work we're no worse off than we are now
3. (Aaron paraphrasing) Include the simple security patch to reject
logins if web.xml has security settings but the Geronimo plan is not
provided or does not have security settings. The proposal to change
our Jetty system to use a "default" realm with no users in it has a
higher risk of breaking something (plus, it's not ready).
4. Tag and cut a set of binaries tonight and start a TCK run
Alan thought we should TCK and release the build that Matt made last
night. I thought that we should integrate the changes above and TCK
and release that.
At the end of the conversation Matt asked me to summarize the
conversation and said of the 4-step plan above: "you can give it
my +1
and barring core dumps in Java this is it. I'll build tonight and
ask David Blevins to start the TCK on it".
After that John Sisson pointed out that we have not fixed the
issue of
spaces in the names of certain files in the documentation.
I'm not totally clear whether Matt wanted more input or whether he's
made his final decision as release manager, but I would assume
that if
anyone feels strongly that the plan above is a mistake then they
should speak up right away. :)
Thanks,
Aaron
--
Geir Magnusson Jr +1-203-665-6437
[EMAIL PROTECTED]
--
Geir Magnusson Jr +1-203-665-6437
[EMAIL PROTECTED]
