Manual work for generation of LICENSE files

[abhijeet gaikwad <ab...@gmail.com>]

Hi Devs,
While working on the release, faced few issues w.r.t. License.txt:
1. I found that we do not record the license of the third-party libs
anywhere in the repository. There should be some place where we must
add the licenses that are redistributed irrespective of the distro
package. This is because - the developer adding the JAR is the best
person to include the License and; we don't miss on including License
for any JARS being shipped.
2. I found that there is a need to manually change the License files
while creating the release artifacts - license files being different
for all 5 packages we ship(src, all bin packages acc. to hadoop
version - 20,23,1.x,2.x).
3. With so many distro packages it is a confusion as to what all
licenses should be present in the LICENSE.txt present in trunk.

The motive of this mail is to discuss on the points as to how we can
maintain and automate the Licenses.

Jarcec has already filed a JIRA for this purpose -
https://issues.apache.org/jira/browse/SQOOP-437. We can consider the
discussion as a part of this JIRA.

Just to start off:
a. For #1 above I think we can have a single file containing all the
licenses(or one file per license/jar) irrespective of the distro, devs
to keep on  appending/removing licenses in this(these) file(s).
b. For #2, we will be able to automate the license file generation
process - if we have all the licenses somewhere in trunk. Suggestions
would be appreciated on how we can automate this generation using any
framework - last/backup approach being adhoc script.
c. For #3, we can take following approaches:
  i. All the licenses to be kept in the LICENSE.txt present in trunk.
I personally vote for this because it will also solve the problem of
maintaining all the licenses - mentioned in #1 above.
  ii. Keep licenses of jars to be included in src package only.
  iii. Keep the license file as it is now.

Let me know your suggestions. It would appreciated if anyone can
suggest any other options that we can take.

Thanks,
Abhijeet

Re: Manual work for generation of LICENSE files

[abhijeet gaikwad <ab...@gmail.com>]

Hi Jarcec,
Thanks for your reply. Responses inline.

On 27 Aug 2012 11:19, "Jarek Jarcec Cecho" <ja...@apache.org> wrote:
>
> Hi Abhijeet,
> thank you very much for taking up this issue and providing very nice
summary.
>
> I was thinking about doing following when I've filled SQOOP-437:
>
> * Have LICENSE file in trunk that covers all jars directly available in
SVN. My goal here was to have license for all thirdparty packages that user
downloads by "cloning" the repository.

Had given a quick thought to this. I feel license should be included for
the jars that we ship. Maintaining such a license is unnecessary.

> * Have prepared license files for all targets stored somewhere in
repository (probably somewhere in src/).

This seems like keeping redundant licenses.

> * Commands "ant tar" or "ant srctar" would automatically pick up correct
license file based on the build profile and replace the default license
file from trunk.
>
> I believe that this procedure will help a bit with distributing correct
license files to appropriate tarballs, however it probably will not remove
the burden from release manager to verify their correct content

Agree.

> What do you think?

I have no strong objections against any of your suggestions. But I feel,
let us wait for another day on this thread and hear from the community.
If no reply by tomorrow - let us go with your suggestions. Will upload the
patch on the JIRA you created.
Thanks!

> Jarcec
>
> On Sat, Aug 25, 2012 at 01:00:40PM +0530, abhijeet gaikwad wrote:
> > Hi Devs,
> > While working on the release, faced few issues w.r.t. License.txt:
> > 1. I found that we do not record the license of the third-party libs
> > anywhere in the repository. There should be some place where we must
> > add the licenses that are redistributed irrespective of the distro
> > package. This is because - the developer adding the JAR is the best
> > person to include the License and; we don't miss on including License
> > for any JARS being shipped.
> > 2. I found that there is a need to manually change the License files
> > while creating the release artifacts - license files being different
> > for all 5 packages we ship(src, all bin packages acc. to hadoop
> > version - 20,23,1.x,2.x).
> > 3. With so many distro packages it is a confusion as to what all
> > licenses should be present in the LICENSE.txt present in trunk.
> >
> > The motive of this mail is to discuss on the points as to how we can
> > maintain and automate the Licenses.
> >
> > Jarcec has already filed a JIRA for this purpose -
> > https://issues.apache.org/jira/browse/SQOOP-437. We can consider the
> > discussion as a part of this JIRA.
> >
> > Just to start off:
> > a. For #1 above I think we can have a single file containing all the
> > licenses(or one file per license/jar) irrespective of the distro, devs
> > to keep on  appending/removing licenses in this(these) file(s).
> > b. For #2, we will be able to automate the license file generation
> > process - if we have all the licenses somewhere in trunk. Suggestions
> > would be appreciated on how we can automate this generation using any
> > framework - last/backup approach being adhoc script.
> > c. For #3, we can take following approaches:
> >   i. All the licenses to be kept in the LICENSE.txt present in trunk.
> > I personally vote for this because it will also solve the problem of
> > maintaining all the licenses - mentioned in #1 above.
> >   ii. Keep licenses of jars to be included in src package only.
> >   iii. Keep the license file as it is now.
> >
> > Let me know your suggestions. It would appreciated if anyone can
> > suggest any other options that we can take.
> >
> > Thanks,
> > Abhijeet

Re: Manual work for generation of LICENSE files

[Jarek Jarcec Cecho <ja...@apache.org>]

Hi Abhijeet,
thank you very much for taking up this issue and providing very nice summary.

I was thinking about doing following when I've filled SQOOP-437:

* Have LICENSE file in trunk that covers all jars directly available in SVN. My goal here was to have license for all thirdparty packages that user downloads by "cloning" the repository.
* Have prepared license files for all targets stored somewhere in repository (probably somewhere in src/).
* Commands "ant tar" or "ant srctar" would automatically pick up correct license file based on the build profile and replace the default license file from trunk.

I believe that this procedure will help a bit with distributing correct license files to appropriate tarballs, however it probably will not remove the burden from release manager to verify their correct content.

What do you think?

Jarcec

On Sat, Aug 25, 2012 at 01:00:40PM +0530, abhijeet gaikwad wrote:
> Hi Devs,
> While working on the release, faced few issues w.r.t. License.txt:
> 1. I found that we do not record the license of the third-party libs
> anywhere in the repository. There should be some place where we must
> add the licenses that are redistributed irrespective of the distro
> package. This is because - the developer adding the JAR is the best
> person to include the License and; we don't miss on including License
> for any JARS being shipped.
> 2. I found that there is a need to manually change the License files
> while creating the release artifacts - license files being different
> for all 5 packages we ship(src, all bin packages acc. to hadoop
> version - 20,23,1.x,2.x).
> 3. With so many distro packages it is a confusion as to what all
> licenses should be present in the LICENSE.txt present in trunk.
> 
> The motive of this mail is to discuss on the points as to how we can
> maintain and automate the Licenses.
> 
> Jarcec has already filed a JIRA for this purpose -
> https://issues.apache.org/jira/browse/SQOOP-437. We can consider the
> discussion as a part of this JIRA.
> 
> Just to start off:
> a. For #1 above I think we can have a single file containing all the
> licenses(or one file per license/jar) irrespective of the distro, devs
> to keep on  appending/removing licenses in this(these) file(s).
> b. For #2, we will be able to automate the license file generation
> process - if we have all the licenses somewhere in trunk. Suggestions
> would be appreciated on how we can automate this generation using any
> framework - last/backup approach being adhoc script.
> c. For #3, we can take following approaches:
>   i. All the licenses to be kept in the LICENSE.txt present in trunk.
> I personally vote for this because it will also solve the problem of
> maintaining all the licenses - mentioned in #1 above.
>   ii. Keep licenses of jars to be included in src package only.
>   iii. Keep the license file as it is now.
> 
> Let me know your suggestions. It would appreciated if anyone can
> suggest any other options that we can take.
> 
> Thanks,
> Abhijeet