Willem:
> I don't want give the customer false assumption that we do regular patch
> release for the every major release of Camel from now on. I just want
> user to move on to the new major release of Camel and we can promise
> them we keep fixing the bugs in the trunk if they report it.
Why not? In what way false assumptions? Personally I would like to see
more patch releases out (3rd digit) and less minor ones (2nd digit). The
benefit and opportunity for the community is huge. There are a bunch of
other projects out there, many oss (like SwitchYard from RH that was
presented at CamelOne, ALv2 licensed too) that would benefit from
maintained branches. We shouldn't force them to upgrade.
Our user base is growing and we need to consider other projects too, not
only paying customers. If you want we can discuss general guidelines,
like how many versions back to support, release cycles etc. It would be
a great thing to commit as a community to a certain level of support.
Hadrian
On 07/06/2011 10:14 PM, Willem Jiang wrote:
On 7/6/11 1:55 PM, Christian Schneider wrote:
Hi Willem,
I can explain a bit from my experience as a user of Camel and CXF at my
former employer regarding patch releases.
We updated our stack every 6 to 8 months. With CXF this mostly simply
worked. When some bug was detected we could create an issue and help fix
it. It went into the next patch release and we could
update to this one instead of the feature release.
With Camel it was different. Every new release had a lot of new features
and changes. So we almost every time found a bug in the release that
prevented us to switch or that was a problem in production that needed
to be addressed. What went very well in Camel was reporting and fixing
bugs. I think Camel is probably the project I used where fixes to bugs
were made fastest. The problem was that the fix was only on trunk. Then
later it was incorporated into the new version. So my first strategy was
to update to the most current camel release when a bug was found and
fixed. The problem was that we almost every time found another problem
in the new release.
So what I did in the end was building our own release with the patches
to the bugs we had. This worked very well but not every customer wants
to do this.
Camel is involved the very fast if you compare it with CXF.
There could be more than 300 issues resolved in one major release of
Camel. CXF is more stable after more than 5 years development, I can
easily merge my patch for CXF trunk to 2.4.x and 2.3.x without changing
anything. But if you merge the bug patch from Camel 2.8-SNAPSHOT to
Camel 2.6.x, you may hit lots of conflicts.
I don't want give the customer false assumption that we do regular patch
release for the every major release of Camel from now on. I just want
user to move on to the new major release of Camel and we can promise
them we keep fixing the bugs in the trunk if they report it.
Patch releases like 2.7.3 give the customer the fixes they need without
the breaking changes that cause new bugs. So from a customer standpoint
patch releases are very valueable. Of course they make life for us more
complicated so I think we should mostly only support one patch release
and one feature release at the same time.
I agree we should do the the patch release for any security reason,
because we take this responsibility without any condition.
Of course a company like Fuse
or Talend can also do patch releases on their own but I think it is
better to have these releases at apache so it is transparent to the
customer what is in each release and he has no fear of vendor lock in.
Current we don't have any other rule for the patch release of 2.7.x,
even 2.8.x. I bring this question out is I don't want to give the user a
false promise that 2.7.x patch release has all the bug patches those we
found in the new major release of 2.8.0 etc.
Christian
Am 06.07.2011 04:09, schrieb Willem Jiang:
Hi Hadrian,
I think we are ready for the Camel 2.8.0 release.
But I'm not sure why you are still planing to do the patch release for
the 2.7.x as we never do this kind small patch release unless it
relates to a serious security issue before.
Can we just let the people move on to Camel 2.8.0 instead of confusing
about what's difference between the Camel 2.8.0 and Camel 2.7.3 ?
On 7/5/11 12:11 PM, Hadrian Zbarcea wrote:
Karaf 2.2.2 is now available and Willem did the upgrade. I think we can
get ready to start the release. Are there any other issues that must go
into 2.8.0?
I would also build a 2.7.3 at the same time, there are a few fixes and
improvements, including some around xmlsecurity.
Thoughts?
Hadrian
On 06/30/2011 11:07 PM, Willem Jiang wrote:
Hi,
I just applied the patch into trunk.
On 7/1/11 12:36 AM, Donald Whytock wrote:
https://issues.apache.org/jira/browse/CAMEL-3948
On Thu, Jun 30, 2011 at 12:35 PM, Donald Whytock<dwhyt...@gmail.com>
wrote:
Just a reminder...CAMEL-3948 is marked as fixed, but the current
trunk
still needs my final patch.
Don
On Thu, Jun 30, 2011 at 2:46 AM, Jean-Baptiste
Onofré<j...@nanthrax.net> wrote:
Hi Claus,
Regarding Karaf 2.2.2, I've released OPS4J dependencies yesterday.
Jamie
will cut off the release this afternoon.
Regards
JB
On 06/30/2011 08:31 AM, Claus Ibsen wrote:
Hi
Okay the JIRA roadmap for Camel 2.8 seems good now. There is 2
open
tickets.
- https://issues.apache.org/jira/browse/CAMEL-3774
- https://issues.apache.org/jira/browse/CAMEL-4144
CAMEL-3774 is about generating the manual and is assigned to
Hadrian.
CAMEL-4144 is about upgrading to Karaf 2.2.2. That release is in
progress.
So by good chance we should be able to pickup that version when
its
released.
Alternatively we can stick to Karaf 2.2.1 which works fine.
(CAMEL-4144 is about some maven validate goal that would require
Karaf
2.2.2 to pickup a fix in Karaf, but running Camel in Karaf is
absolutely fine)
The CI servers also seems good. Although they tend to run out of
memory at the end, such as when testing the examples. But those
are
the last piece of the build, and thus all components tests fine.
I suggest that when Karaf 2.2.2 is out we git it a few spins on
the CI
servers and then start cutting the Camel 2.8 release. Would be
good to
get it out before the summer vacation starts. As well its more
than 3
months since Camel 2.7 was released.
On Mon, Jun 27, 2011 at 9:35 AM, Claus
Ibsen<claus.ib...@gmail.com>
wrote:
Hi
Okay we should really start focusing on getting the last tickets
which
has been assigned for 2.8 release done now.
There is about 350 tickets on the roadmap, so its going to be the
biggest release, since 2.0 went GA.
So please take a look at your assigned tickets and get them
done, or
move them for 2.9.
Then keep eyes on CI servers and help fix any test failures,
so we
have green builds.
The summer vacation period is approaching so we should IMHO get
the
2.8 release out early next month if possible.
On Thu, Jun 9, 2011 at 2:47 PM, Hadrian
Zbarcea<hzbar...@gmail.com>
wrote:
Hi,
I would propose starting to close down and prepare for the 2.8.0
release
in
2-3 weeks. There are already 282 issues for 2.8.0 and chances
are
will
be
over 300 by the release time, probably setting a new record.
As of now there are 17 issues unresolved, a few of them almost
done, so
by
next week I assume there'll be significantly less. I would
suggest
shifting
the focus from adding new features to stabilizing the build. If
there
are
any issues you know of that you think absolutely must be in
2.8.0
please
shout and ask for help if needed (especially non committers
subscribing
to
this list).
Thoughts?
Hadrian
--
Claus Ibsen
-----------------
FuseSource
Email: cib...@fusesource.com
Web: http://fusesource.com
Twitter: davsclaus, fusenews
Blog: http://davsclaus.blogspot.com/
Author of Camel in Action: http://www.manning.com/ibsen/
