I don't see why requirement to commit changes to the master can or
should be relaxed. The master is the code line that is and should be
maintained by the community. The repository is not a collection of
individual branches where I maintain my own branch and you maintain your
own (it is an extreme, but that what the proposal to allow submissions
to release-3 without a submission to the master eventually leads to). If
a contributor does not want to make changes on the master, he/she can
continue development on a private fork and release it (not as official
Apache release). At this point there will be no guarantee that the
changes would ever make into the Apache release (whether on master or
any other branch). IMO, once the community moves to the next major
release, all feature and not critical bug fixes should go into the
master. If there is a critical bug fix, it must be fixed in the master
and if it is critical enough, backported to whatever branch a
contributor is willing to backport it to.
If there are code changes that only apply to release-3 branch and do not
apply to the master (for whatever reason), this should be spelled out in
JIRA and be considered an exception, not a rule.
Thank you,
Vlad
On 8/25/17 12:37, Pramod Immaneni wrote:
On Fri, Aug 25, 2017 at 10:16 AM, Thomas Weise <t...@apache.org> wrote:
There is obviously disagreement with regard to version reset and you have
already voted for 4.0.
Yes and my position is still the same. I think there is a general consensus
with 4.0, the additional concern/ask from community I believe, is because
of the divergence as things go forward, to not burden the contributor from
having to submit the PR to both master and release-3 and to relax the
restriction by allowing submissions to go into release-3 without requiring
they go into master. This would be different from what we do today.
I think we can accomodate this for the following reasons. For the most part
fixes and updates to existing operators/code would be cherry pickable onto
master as is, with simple conflict resolution or with minor changes on top
such as package name changes. If the contributor does not want to do the
work the committer or the community can step in and do this at the same
time when the original PR is merged or asynchronously if more time is
needed so as to not block the original PR. We will need a way to track
pending items and possibly we could do this in JIRA itself and additionally
not close the JIRA till it is also fixed in master. The ones that are not
trivial to port would be few in my opinion because as things go forward
majority of the contributions will just go to master and for these few we
can again rely on the community for help. Also, with the prevailing
sentiment to do additional refactoring and cleanups in 4.x there may be no
reason to port some of the changes in release-3 as they may no longer be
applicable to 4.x.
A larger issue though is the attempt to stop any other proposal without
valid reason.
If community members are willing to invest their efforts to improve the
project and provide the justification
for the changes then those that otherwise don't participate in discussions
or initiatives should not seek to prevent changes from happening.
A project that is stuck in the past won't attract volunteers to work on it
and users to adopt it.
There are different users who have adopted the project at different stages
or versions and have made investments on their side adopting the project
and as the project continues to grow the number will only increase. Changes
against the grain like these, affect them and we cannot trivialize it as
merely a few lines code change or a pom.xml change. That may be a new
development cycle for them and the associated cost. There may be
documentation they need to change for their developers. We have to listen
to community opinion and chart the best course forward.
Thanks
Thomas
On Fri, Aug 25, 2017 at 9:57 AM, Pramod Immaneni <pra...@datatorrent.com>
wrote:
No, concerns for the changing the project name and version remain the
same.
I think the current version numbering train and name are fine and prefer
we
move forward and not backward by resetting things back to 1.x, which I
believe is not accurate for the project. The name change is unnecessary,
the name isn't broken that it needs to be fixed, it does not buy us much
and causes unnecessary disruption for people who are already used to
and malhar is already known out there. It is equivalent to asking for
apex-core to be changed to apex-engine, engine is probably a better name
but is it worth making the change, probably not, if it is not broke why
fix
it.
Thanks
On Fri, Aug 25, 2017 at 9:01 AM, Vlad Rozov <vro...@apache.org> wrote:
How do we move from here? If all the concerns regarding version and
artifactId change are addressed we should move forward with the vote,
if
not, it will be good to raise them here rather than in the voting
thread.
Thank you,
Vlad
On 8/24/17 10:26, Thomas Weise wrote:
On Thu, Aug 24, 2017 at 9:42 AM, Amol Kekre <a...@datatorrent.com>
wrote:
In terms of rebasing versions, there is no urgency in mimic-ing some
of
the
other projects. Apex has already be been versioned.
There is an expectation users have for a version number, which is
different
for 3.x or 1.x or 0.x. Apex library maturity is nowhere near 3.x. That
was
already discussed.
What functional gain do
we have by changing versions, names? Functionality wise Apex users do
not
gain anything. With regards to bumping to 4.X, we should wait for a
proposal/plan for a new functional api.
Addition of such API does not require major version change. New API
have
been added and no major version change was done. Major version change
is
for backward incompatible changes.
Examples:
- rename packages
- remove deprecated code
- relocate operators that were not designed for production use
- change to functionality of operators
There is an illusion of backward compatibility (which does not exist
today). That cannot be used as justification to not make changes.
On Wed, Aug 23, 2017 at 10:26 AM, Vlad Rozov <vro...@apache.org>
wrote:
Please see my comments in-line.
Thank you,
Vlad
On 8/23/17 09:11, Pramod Immaneni wrote:
That is not accurate, I have mentioned and probably others as well
that
changing the name of the project would be disruptive to users.
Users
are
used to using the malhar project and its artifacts a certain way
and
this
would cause them immediate confusion followed by consternation and
then
changes that could extend beyond their application such as
documentation
etc.
Changing the name is as disruptive to users as changing minor/patch
version. I don't see a big difference in changing one line in
pom.xml
(version) vs changing 2 lines (version and artifact). There is a
bigger
change/disruption that does IMO require major version change and
renaming
project to use the single brand (Apache Apex) at the same time is
beneficial both to the project and users. Changing package and major
version will impact documentation in much bigger way compared to
changing
artifactId.
Second the project has been around for quite some time and has
reached a
version 3.x, the second part of the proposed change is to reset it
back
to
1.0-SNAPSHOT. I don't think that is accurate for the project and the
maturity it would portray to the users. Not to get subjective but
there
are
operators in malhar that are best of the breed when it comes to
streaming
functionality they achieve.
There are many Apache projects that were around much longer than
malhar
and have not yet reached 3.x version even though they are also used
in
production and are considered more stable. Number of evolving
packages
and
interfaces in malhar do not qualify it for 3.x or 4.x. IMO, version
must
be
driven by the engineering/community, not by the marketing.
Third think about all the changes it would need, code, project
infrastructure such as github repo and jira project, documentation,
website
etc and the time all the developers have to spend to adapt to this.
Wouldn't we want to spend this time doing something more
productive.
I don't think it is as drastic as it looks to be. It was done in a
past
and is supported by all tools involved.
I would think changing a project name and resetting the version is a
big
deal and should be done if there something big to gain for the
project
by
doing this. What is the big gain we achieve to justify all this
consternation? If we want to increase adoption, one of the things
we
need
to do is to provide users with a platform that behaves in an
expected
and
stable manner.
It will be good to provide details why is it "a big deal". Why
changing
groupId was not a big deal and changing artifactId is a big deal?
I completely agree with the increasing adoption, but it comes from
the
quality, not from the quantity and whether version is 1.x, 3.x or
4.x
does
not change the quality of the library.
Thanks
On Wed, Aug 23, 2017 at 8:09 AM Vlad Rozov <vro...@apache.org>
wrote:
All -1 are technically void at this point as justification given
are
why
project may continue without modifications and not why the
modification
must not be done. Whether we proceed with the vote or with the
discussion, arguments should be what are pros and cons of a code
change,
not that the project may continue without them. The same should
apply
not only to the current set of changes, but to all future
discussions.
Thank you,
Vlad
On 8/23/17 06:54, Thomas Weise wrote:
The discussion already took place [1]. There are two options under
vote
out
of that discussion and for the first option there is a single -1.
Use
of
-1
during voting (and veto on PR) when not showing up during the
preceding
discussion is problematic.
Thomas
[1] https://lists.apache.org/thread.html/
bd1db8a2d01e23b0c0ab98a785f6ee
9492a1ac9e52d422568a46e5f3@%3Cdev.apex.apache.org%3E
On Wed, Aug 23, 2017 at 1:58 AM, Justin Mclean <
jus...@classsoftware.com
wrote:
Hi,
Votes are only valid on code modifications with a reason. [1]
However it looks to me that there’s not consensus and which way
forward
is
best I would suggest cancelling the vote and having a discussion
of
the
benefit or not of making the change.
Thanks,
Justin
1. https://www.apache.org/foundation/voting.html
Thank you,
Vlad
Thank you,
Vlad
Thank you,
Vlad
Thank you,
Vlad
