cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rajani Karuturi <Rajani.Karut...@citrix.com>
Subject Re: Review Request 23192: Adding Readme and run checkbox at the end of the installation. Also installing mysql connector
Date Mon, 14 Jul 2014 03:53:02 GMT
Hi Hugo,

Does this make sense? Do you still think it should be part of noredist profile?

~Rajani



On 09-Jul-2014, at 10:44 am, Rajani Karuturi <Rajani.Karuturi@citrix.com<mailto:Rajani.Karuturi@citrix.com>>
wrote:

Hi Hugo,
Even in case of windows installer, mysql-connector-java is mentioned in a [spec] file. only
difference being, it also mentions the download url as windows doesn’t have a package repository
and doesn’t know where to download from.
It will not be part of the installer. The msi only tells where to download the mysql connector
jar from. It doesn’t include the jar itself.

When the user is installing cloudstack using the msi, the jar will be downloaded from the
location mentioned.


[spec] https://reviews.apache.org/r/23192/diff/#5

~Rajani



On 08-Jul-2014, at 7:25 pm, Hugo Trippaers <hugo@trippaers.nl<mailto:hugo@trippaers.nl>>
wrote:

Rajani,

That is why we renamed nonoss to noredist. It has to do with the policy of the Apache Software
Foundation regarding compatible licenses [1]. The problem in this case is subtle, but there
is a difference. The main point in the discussion is so-called system dependencies. We know
we need java, mysql and the mysql-connector-java to be installed before we can install our
packages. This is what the dependency on mysql-connector-java in the spec file indicates.
Creating a package containing those source is actually including the items we consider is
going just a bit further and to a point where we concluded that it is not inline with the
apache policies. So it is perfectly alright to include a popup in the installer that tells
the users that he needs to install mysql, java etc, but we can’t have an installed in the
source that will do that for the user as part of the installer.

What David proposes is a the middle ground where we can keep the installer in the configured
like this in the pom, but it can’t be in that standard build, but only in the build that
know we can’t redistribute (the noredist build). We will never make the artifacts resulting
from that build available, but other entities can make them available. By specifying noredist
on the build command we can be reasonably sure that the builder has been informed that distribution
restrictions might apply on whatever it is he is doing.

Cheers,

Hugo

[1] http://www.apache.org/legal/resolved.html#category-x


On 8 jul. 2014, at 12:57, Rajani Karuturi <Rajani.Karuturi@citrix.com<mailto:Rajani.Karuturi@citrix.com>>
wrote:

I believe this has nothing to do with nonoss. If I understand it right, nonoss modules are
required for cloudstack to work with certain 3rd party stuff.

This one is just a helper to create a windows installer. It chose to use a mvn command to
create the installer instead of shell/bat script. It should work with either nonoss or oss
modules.
This is very similar to cloud.spec[1] we have which is used to create the rpm and which already
has mysql-connector-java already listed as required package.

I think the right way would be to create independent git repos for deb/rpm/windows installers/packages.

In its current state, I don’t see it any different from other helper installer scripts we
already have.

[1] https://git-wip-us.apache.org/repos/asf?p=cloudstack.git;a=blob_plain;f=packaging/centos63/cloud.spec;hb=HEAD

~Rajani



On 08-Jul-2014, at 9:56 am, Damoder Reddy <Damoder.Reddy@citrix.com<mailto:Damoder.Reddy@citrix.com><mailto:Damoder.Reddy@citrix.com>>
wrote:

Ok,

To clarify the windows MSI will not build in the default profile. I have enabled a new profile
"buildw" to build windows MSI installer which explicitly we need to pass similar to nonoss?

Is that sufficient to make it nonoss or still we need to move it under nonoss profile explicitly?

Thanks
Damoder/

-----Original Message-----
From: David Nalley [mailto:david@gnsa.us]
Sent: Thursday, 3 July 2014 9:58 PM
To: dev@cloudstack.apache.org<mailto:dev@cloudstack.apache.org><mailto:dev@cloudstack.apache.org>
Cc: Damoder Reddy; Koushik Das
Subject: Re: Review Request 23192: Adding Readme and run checkbox at the end of the installation.
Also installing mysql connector

On Thu, Jul 3, 2014 at 12:08 PM, Chip Childers <chipchilders@apache.org<mailto:chipchilders@apache.org><mailto:chipchilders@apache.org>>
wrote:
On Thu, Jul 03, 2014 at 03:14:30PM +0000, Leo Simons wrote:
It looks like that maven pom on windows _by default_ downloads and
installs a variety of non-apache-license (and/or non-mit/bsd/variant
license) software. That shouldnąt really happen. The principle is one
of łleast surprise˛: As a user or developer who does not RTFM,
following the default commands/tools/etc, you should end up with a
more-or-less apache-licensed build result (*) that you can
redistribute the result under.

+1


But apache policy is that it is acceptable to provide scripts/build
tools/assistance to help those same users/developers do things that
they want to do. As long as they understand the legal situation they end up in.

I would recommend adding a "nonoss" maven profile that the
developer/user has to explicitly select in order to do those
downloads. As long as that option is described clearly, thatąs then
ok. See

http://svn.apache.org/repos/asf/apr/apr/trunk/README

for an example of how to point out the license situation.

We already have the nonoss profile, so this is a pretty good fix for
the windows build issues noted above IMO.

Damoder/Koushik - please make this change.


Something similar is true by the way (IMHO, but as a project
cloudstack can definitely decide differently), for a possible MSI
script. Making an MSI script that prompts the user whether to
download mysql at the point of install, **clearly pointing out the
license situation** if they choose to do so, seems reasonable, and I
personally would not object to shipping _that_ kind of script as part of an apache source
release.


+1 - that's a reasonable approach as well.  Damoder / Koushik - what
+do
you think about this approach?


I like this approach. We have a number of things that aren't in the 'default' build because
of policy reasons. This is just another of them.



Finally, the _spirit_ behind the apache policies is that there should
be an option to use cloudstack with a license-compatible database
(say, postgres), even if most users will use mysql (just like most
people that use dbm with httpd will use berkely dbm, but you _can_
use something else). Itąs perhaps unfortunate that this isnąt
supported, but thatąs not apache policy, and given the license
situation of other system dependencies, I can imagine no-one here wants to make it a priority.

Yeah, that would be nice...  but somebody would have to decide that
they want to do that.



cheers,


Leo

PS: IANAL, but, a lot of this discussion is a bit beyond legal, and
is about choice/policy, and the policy is supposed to be based on
common sense much more than license stuff tends to be :)

Agreed - this is about policy not legality.





Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message