Ignite-1.5 Release

[dsetrakyan]

On Mon, Nov 30, 2015 at 9:17 AM, Raul Kripalani <[hidden email]> wrote:

> On Mon, Nov 30, 2015 at 5:05 PM, Yakov Zhdanov <[hidden email]>
> wrote:
>
> > Raul, every build submitted for vote is RC (so, on first iteration this
> > will be 1.5.0-EA-RC1). Once accepted RC gets stripped off and we get
> > official release. My idea is to have EA available via maven and from the
> > website.
> >
>
> Aha, I see. I'm not sure I like the idea of pushing the EA to Maven Central
> and making it live there perpetually. After 3 years, we would still have an
> 1.5.0.EA there.
> Red Hat is a company that uses EAs quite often. They have a separate Maven
> repo for them, from which they prune the EAs once they become GA.
>
> I prefer that we publish a RC1 to a staging repo and inform our users via
> mailing list. All they'd have to do to take 1.5.0-RC1 for a spin is add the
> staging repo to their pom.xml.
>

Raul, personally, I am more in favor of adding the EA to Maven. Having it
there forever does not really scare me. It will be a lot easier for the
community to take the release for a spin without having some special
instructions. Moreover, I imagine that many users will jump on it as well,
as it sounds like the release is pretty close to the production-ready state.



>
> *Raúl Kripalani*
> PMC & Committer @ Apache Ignite, Apache Camel | Integration, Big Data and
> Messaging Engineer
> http://about.me/raulkripalani | http://www.linkedin.com/in/raulkripalani
> http://blog.raulkr.net | twitter: @raulvk
>

[Raul Kripalani]

On Mon, Nov 30, 2015 at 5:23 PM, Yakov Zhdanov <[hidden email]> wrote:

> 1. I see nothing wrong in having EA perpetually available. I would even
> prefer all vendors to keep all EA versions available just for tracking and
> history purposes.
>

We have opposite views here. To me, early access versions are ephemeral.
They are only legit until the final version is released. By doing that, you
run the risk that "forgetful" users who once tried an EA, keep using it
forever even after GA is released.

In other words, what you are proposing to do – to push an EA to Maven
Central – is equivalent to Microsoft packaging Windows 11 Early Access in a
box and putting it on sale forever. And keeping it on sale even after
Windows 11 GA, Windows 12 GA, Windows 13 GA (Americans would probably skip
that :)), etc. are released decades after.

No company or project does that, to my knowledge. I would consider it an
anti-pattern.

Regards,

*Raúl Kripalani*
PMC & Committer @ Apache Ignite, Apache Camel | Integration, Big Data and
Messaging Engineer
http://about.me/raulkripalani | http://www.linkedin.com/in/raulkripalani
http://blog.raulkr.net | twitter: @raulvk

[Sergey Kozlov]

Hi

Some companies provide EA releases in maven:

https://oss.sonatype.org/content/repositories/releases/com/hazelcast/hazelcast/3.6-EA/


On Mon, Nov 30, 2015 at 9:39 PM, Raul Kripalani <[hidden email]> wrote:

> On Mon, Nov 30, 2015 at 5:23 PM, Yakov Zhdanov <[hidden email]>
> wrote:
>
> > 1. I see nothing wrong in having EA perpetually available. I would even
> > prefer all vendors to keep all EA versions available just for tracking
> and
> > history purposes.
> >
>
> We have opposite views here. To me, early access versions are ephemeral.
> They are only legit until the final version is released. By doing that, you
> run the risk that "forgetful" users who once tried an EA, keep using it
> forever even after GA is released.
>
> In other words, what you are proposing to do – to push an EA to Maven
> Central – is equivalent to Microsoft packaging Windows 11 Early Access in a
> box and putting it on sale forever. And keeping it on sale even after
> Windows 11 GA, Windows 12 GA, Windows 13 GA (Americans would probably skip
> that :)), etc. are released decades after.
>
> No company or project does that, to my knowledge. I would consider it an
> anti-pattern.
>
> Regards,
>
> *Raúl Kripalani*
> PMC & Committer @ Apache Ignite, Apache Camel | Integration, Big Data and
> Messaging Engineer
> http://about.me/raulkripalani | http://www.linkedin.com/in/raulkripalani
> http://blog.raulkr.net | twitter: @raulvk
>

--
Sergey Kozlov

[Raul Kripalani-2]

Any others? I need some more convincing :)

That Hazelcast does it doesn't mean that it's correct SW Engineering
practice. Perhaps they don't have a public Maven repo (probably that's the
reason) where they can publish their EA in a staging area. The ASF does
(advantage) and therefore we can use this facility.

Moreover that nomenclature is horrible. It ties them up to a single EA.
What if it comes out with a blocking bug that impedes testing? They'd need
to release a second EA, and that's not taken into account.
On 30 Nov 2015 18:50, "Sergey Kozlov" <[hidden email]> wrote:

> Hi
>
> Some companies provide EA releases in maven:
>
>
> https://oss.sonatype.org/content/repositories/releases/com/hazelcast/hazelcast/3.6-EA/
>
>
> On Mon, Nov 30, 2015 at 9:39 PM, Raul Kripalani <[hidden email]> wrote:
>
> > On Mon, Nov 30, 2015 at 5:23 PM, Yakov Zhdanov <[hidden email]>
> > wrote:
> >
> > > 1. I see nothing wrong in having EA perpetually available. I would even
> > > prefer all vendors to keep all EA versions available just for tracking
> > and
> > > history purposes.
> > >
> >
> > We have opposite views here. To me, early access versions are ephemeral.
> > They are only legit until the final version is released. By doing that,
> you
> > run the risk that "forgetful" users who once tried an EA, keep using it
> > forever even after GA is released.
> >
> > In other words, what you are proposing to do – to push an EA to Maven
> > Central – is equivalent to Microsoft packaging Windows 11 Early Access
> in a
> > box and putting it on sale forever. And keeping it on sale even after
> > Windows 11 GA, Windows 12 GA, Windows 13 GA (Americans would probably
> skip
> > that :)), etc. are released decades after.
> >
> > No company or project does that, to my knowledge. I would consider it an
> > anti-pattern.
> >
> > Regards,
> >
> > *Raúl Kripalani*
> > PMC & Committer @ Apache Ignite, Apache Camel | Integration, Big Data and
> > Messaging Engineer
> > http://about.me/raulkripalani | http://www.linkedin.com/in/raulkripalani
> > http://blog.raulkr.net | twitter: @raulvk
> >
>
>
>
> --
> Sergey Kozlov
>

[yzhdanov]

Raul, I don't see anything wrong with this approach. EA2 can follow EA.
Versioning of EAs does not matter too much to me :)

You can download EAP build of IntellijIdea, although Idea 14 is available
now for more than 1 year already.

Btw, here is one more example of keeping EA versions in repo -
http://mvnrepository.com/artifact/com.sun.jersey/jersey-client

Hibernate seems to keep all versions in maven -
http://mvnrepository.com/artifact/org.hibernate/hibernate-core

Personally I do not insist on releasing this in maven, but I don't think
this is anything extraordinary.

Guys, any more opinions?

--Yakov

2015-11-30 22:12 GMT+03:00 Raul Kripalani <[hidden email]>:

> Any others? I need some more convincing :)
>
> That Hazelcast does it doesn't mean that it's correct SW Engineering
> practice. Perhaps they don't have a public Maven repo (probably that's the
> reason) where they can publish their EA in a staging area. The ASF does
> (advantage) and therefore we can use this facility.
>
> Moreover that nomenclature is horrible. It ties them up to a single EA.
> What if it comes out with a blocking bug that impedes testing? They'd need
> to release a second EA, and that's not taken into account.
> On 30 Nov 2015 18:50, "Sergey Kozlov" <[hidden email]> wrote:
>
> > Hi
> >
> > Some companies provide EA releases in maven:
> >
> >
> >
> https://oss.sonatype.org/content/repositories/releases/com/hazelcast/hazelcast/3.6-EA/
> >
> >
> > On Mon, Nov 30, 2015 at 9:39 PM, Raul Kripalani <[hidden email]>
> wrote:
> >
> > > On Mon, Nov 30, 2015 at 5:23 PM, Yakov Zhdanov <[hidden email]>
> > > wrote:
> > >
> > > > 1. I see nothing wrong in having EA perpetually available. I would
> even
> > > > prefer all vendors to keep all EA versions available just for
> tracking
> > > and
> > > > history purposes.
> > > >
> > >
> > > We have opposite views here. To me, early access versions are
> ephemeral.
> > > They are only legit until the final version is released. By doing that,
> > you
> > > run the risk that "forgetful" users who once tried an EA, keep using it
> > > forever even after GA is released.
> > >
> > > In other words, what you are proposing to do – to push an EA to Maven
> > > Central – is equivalent to Microsoft packaging Windows 11 Early Access
> > in a
> > > box and putting it on sale forever. And keeping it on sale even after
> > > Windows 11 GA, Windows 12 GA, Windows 13 GA (Americans would probably
> > skip
> > > that :)), etc. are released decades after.
> > >
> > > No company or project does that, to my knowledge. I would consider it
> an
> > > anti-pattern.
> > >
> > > Regards,
> > >
> > > *Raúl Kripalani*
> > > PMC & Committer @ Apache Ignite, Apache Camel | Integration, Big Data
> and
> > > Messaging Engineer
> > > http://about.me/raulkripalani |
> http://www.linkedin.com/in/raulkripalani
> > > http://blog.raulkr.net | twitter: @raulvk
> > >
> >
> >
> >
> > --
> > Sergey Kozlov
> >
>

[Raul Kripalani]

On Mon, Nov 30, 2015 at 8:41 PM, Yakov Zhdanov <[hidden email]> wrote:

> Raul, I don't see anything wrong with this approach. EA2 can follow EA.
> Versioning of EAs does not matter too much to me :)
>

If you are wanting to push EAs to Central, then we must have a predictable
versioning scheme. Publishing to Central makes the release perpetual. So
EA2 after EA is not predictable. If you insist on this idea, I would use
EA-01, EA-02, EA-03, etc.

Let's design things out well and avoid being haphazard.


> You can download EAP build of IntellijIdea, although Idea 14 is available
> now for more than 1 year already.
>

You know for a fact that IntelliJ's EAPs are removed once they are no
longer relevant. Because they own their file server and can remove them
whenever they want. That's why only 15 EAP and 14.1 EAP are available.

Haven't you thought why 13 EAP, 12 EAP, etc. are no longer available?


> Btw, here is one more example of keeping EA versions in repo -
> http://mvnrepository.com/artifact/com.sun.jersey/jersey-client
>

You are talking about releases from 2012 and older. Haven't you realised
why they stopped doing that? Please don't bend reality to make it fit your
arguments.

Hibernate seems to keep all versions in maven -
> http://mvnrepository.com/artifact/org.hibernate/hibernate-core
>

Yes, and they publish Betas, Service Packs, Release Candidates, etc. They
have a complex, pragmatic and well-defined lifecycle.
We are just talking about EAs in a very particular situation. It's going to
be extremely random to look back and see:

Ignite 1.2.0
Ignite 1.3.0
Ignite 1.4.0
Ignite 1.5.0-EA
Ignite 1.5.0
Ignite 1.6.0

If we want to define a full lifecycle with public milestones, I'm all up
for it. It does create lots of complexity and release work, though. But
publishing an EA to Central just because in this particular release we
changed something big, to me it's a NO-NO.


> Personally I do not insist on releasing this in maven, but I don't think
> this is anything extraordinary.
>
> Guys, any more opinions?
>

I would like to hear the opinions of people outside of GG. @Cos, @Brane?

And please Yakov, let's respect the community and don't take decisions by
ourself without consensus and without giving people the opportunity to
speak their mind.

*Raúl Kripalani*
PMC & Committer @ Apache Ignite, Apache Camel | Integration, Big Data and
Messaging Engineer
http://about.me/raulkripalani | http://www.linkedin.com/in/raulkripalani
http://blog.raulkr.net | twitter: @raulvk

[dsetrakyan]

Guys, there is obviously a difference of opinions here. If I can make a
suggestion, let’s agree on the versioning scheme and proceed. I like Raul’s
suggestion for EA-01/02/03/etc… If there are no objections, let’s follow
this scheme going forward.

D.

On Mon, Nov 30, 2015 at 3:11 PM, Raul Kripalani <[hidden email]> wrote:

> On Mon, Nov 30, 2015 at 8:41 PM, Yakov Zhdanov <[hidden email]>
> wrote:
>
> > Raul, I don't see anything wrong with this approach. EA2 can follow EA.
> > Versioning of EAs does not matter too much to me :)
> >
>
> If you are wanting to push EAs to Central, then we must have a predictable
> versioning scheme. Publishing to Central makes the release perpetual. So
> EA2 after EA is not predictable. If you insist on this idea, I would use
> EA-01, EA-02, EA-03, etc.
>
> Let's design things out well and avoid being haphazard.
>
>
> > You can download EAP build of IntellijIdea, although Idea 14 is available
> > now for more than 1 year already.
> >
>
> You know for a fact that IntelliJ's EAPs are removed once they are no
> longer relevant. Because they own their file server and can remove them
> whenever they want. That's why only 15 EAP and 14.1 EAP are available.
>
> Haven't you thought why 13 EAP, 12 EAP, etc. are no longer available?
>
>
> > Btw, here is one more example of keeping EA versions in repo -
> > http://mvnrepository.com/artifact/com.sun.jersey/jersey-client
> >
>
> You are talking about releases from 2012 and older. Haven't you realised
> why they stopped doing that? Please don't bend reality to make it fit your
> arguments.
>
> Hibernate seems to keep all versions in maven -
> > http://mvnrepository.com/artifact/org.hibernate/hibernate-core
> >
>
> Yes, and they publish Betas, Service Packs, Release Candidates, etc. They
> have a complex, pragmatic and well-defined lifecycle.
> We are just talking about EAs in a very particular situation. It's going to
> be extremely random to look back and see:
>
> Ignite 1.2.0
> Ignite 1.3.0
> Ignite 1.4.0
> Ignite 1.5.0-EA
> Ignite 1.5.0
> Ignite 1.6.0
>
> If we want to define a full lifecycle with public milestones, I'm all up
> for it. It does create lots of complexity and release work, though. But
> publishing an EA to Central just because in this particular release we
> changed something big, to me it's a NO-NO.
>
>
> > Personally I do not insist on releasing this in maven, but I don't think
> > this is anything extraordinary.
> >
> > Guys, any more opinions?
> >
>
> I would like to hear the opinions of people outside of GG. @Cos, @Brane?
>
> And please Yakov, let's respect the community and don't take decisions by
> ourself without consensus and without giving people the opportunity to
> speak their mind.
>
> *Raúl Kripalani*
> PMC & Committer @ Apache Ignite, Apache Camel | Integration, Big Data and
> Messaging Engineer
> http://about.me/raulkripalani | http://www.linkedin.com/in/raulkripalani
> http://blog.raulkr.net | twitter: @raulvk
>

[Ognen Duzlevski]

On Mon, Nov 30, 2015 at 12:39 PM, Raul Kripalani <[hidden email]> wrote:

>
> We have opposite views here. To me, early access versions are ephemeral.
> They are only legit until the final version is released. By doing that, you
> run the risk that "forgetful" users who once tried an EA, keep using it
> forever even after GA is released.
>

Apache Ignite is a product for professionals - engineers etc. If someone
"forgets" that their production is running the EA version, well, they need
to look for a different profession ;). Comparing it to Windows 12 EA is
silly, IMHO, since most people using Windows are technically not at the
same level, they are far below.

My $.02.