Ok...if we are in favor of an August M5, then I am cool with this.
I want to point out a little history to understand why there is some
passion in getting this into M4.
My understanding about M(X), is the M stands for Milestone. I believe
that you hit a milestone based on a set of criteria. I not only
understood that we had concensus that this would be a part of milestone
4 and thus fit the criteria, but I was specifically asked by one of the
committers on the project to get it done fast so it can be in M4. I
worked very hard to get this out. It was somewhat disconcerting to me
that this got kicked back to M5 after all of this...as I could have
taken my time.
I would only ask that A) M5 come in a relative short time period...and
more importantly, that B) we decide ahead of time what is in the
milestone, and cut based on the milestone roadmap's requirements of
included options, as opposed by date. Although it would seem that A
conflicts with B, it does not. If we keep our list relatively short, A
will work with B.
Finally, IMHO, there has been too much committer talk and decision on
this subject. I really would love to hear what the community wants in
M4 and M5. The community's voice should be the most important.
Jeff
David Jencks wrote:
I am extremely strongly in favor of only bug fixes for M4 and putting
out M5 in mid august and 1.0 in mid sept.
I'd like to point out that re-branching at this point is essentially
abandoning M4 for M5. I have committed substantial changes to head that
are not yet necessarily completely stable and are also not quite
complete. If we rebranch, we won't be able to get the new M4.5 out
before mid august anyway.
I abandoned my hopes of getting all the work I am putting in head into
M4: after some consideration I decided that it was more important to get
M4 out than my favorite features in, even though I was sure they would
not be destabilizing. There's some difference in that my features have
no discernable impact on the normal user, but still :-)
thanks
david jencks
On Jul 20, 2005, at 6:56 AM, [EMAIL PROTECTED] wrote:
On the Geronimo IRC channel there was talk about the Tomcat/Jetty Picker
not going in M4 because it is now involving more code changes than what
people thought they had agreed to. This was a surprise to me and after
discussion it was proposed that I call for a vote.
Before I do, I thought a little background might be helpful..
Back in the mail thread "Preparation for M4 -- jetty vs tomcat or jetty
and tomcat (two builds)" on 5th of July it was agreed that there
would be
a Tomcat and a Jetty build of Geronimo.
In the mail thread "Wait or not? Respond quick. (M4 -- 24 hour notice of
branch)" on 9th of July, it appears nobody asked to hold off creating the
branch to do the work for the Tomcat / Jetty builds. Maybe it was just
assumed it was going to be simple changes in the branch, or it was
forgotten.
In the mail thread "M4 Status", started by Aaron on 18th July, he said "I
believe Jeff is working on separate plans for Tomcat and Jetty builds, so
we can produce two separate distributions as people seemed to prefer." .
Alan responded "I think that the notion that adding new features into a
QA branch is a bad idea stands, regardless of how simple the changes are
and how simple it is to merge them. It's simply bad form". Alan then
said "I'm not opposed to the what and why. I am opposed to the how."
David Jencks also agreed with Alan in the mail thread.
So it seems that people are unhappy with the "how" as Alan said.
Since it was already agreed that we are to have separate Tomcat and Jetty
builds in M4, that decision should not be questioned and as a reminder
Jeff's changes have the following benefits:
* Less user problems - the previous method of having to edit many
files is
prone to failure, it caught me out many times, and I have seen others get
caught out!
* We don't have to document the M4 way of configuring the web containers
and the M5 way of configuring. This makes the instructions more
complicated and makes it harder for other forms of documentation to stay
relevant (e.g. articles and Aaron's book).
* Documentation does not have to be changed when we reach M5.
* We are seen to be trying to minimise changes that impact configuration
between releases.
Looking back, it appears we branched too early.
I propose that we vote on the "how" with the following options:
a) Merge Jeff's Tomcat/Jetty switch changes into the M4 QA branch
b) Make a new Geronimo M4 QA and OpenEJB M4 QA branches from HEAD
when it is stable.
John
This e-mail message and any attachments may contain confidential,
proprietary or non-public information. This information is intended
solely for the designated recipient(s). If an addressing or transmission
error has misdirected this e-mail, please notify the sender immediately
and destroy this e-mail. Any review, dissemination, use or reliance upon
this information by unintended recipients is prohibited. Any opinions
expressed in this e-mail are those of the author personally.
