Partial redistribution of Intel(R) Distribution for Python* allowed?

[Manuel_Z_]

Hello Community, Hello Intel,

the Intel Simplified Software License seems to be a license which encourages redistribution of derived works.

That's why I'm currently thinking about switching from vanilla numpy+scipy to the ones provided with the Intel(R) Distribution for Python*.

There are a few things that make me wonder.

Question 1a: Can I redistribute the Intel(R) Distribution for Python* only in whole or also in parts?

- The license says: You may use and redistribute the software without modification

- Am I allowed to redistribute individual packages like numpy+scipy in an isolated form?

Question 1b: Can I redistribute the Intel(R) Distribution for Python* in a somewhat statically bundled form?

- My current build-process packs most libs (own ones and all dependencies) into on large tar-File. I consider this to be some kind of "static linking" which for example could violate the LGPL.

- Is the Intel(R) Distribution for Python* (or parts of it) fine with this?

- Would it be Okay to pack numpy+scipy from Intel(R) Distribution for Python* into this tar-File?

Question 2: What about the redist.txt?

- It referes to a End User License Agreement for the Intel(r) Distribution for Python ("Agreement").

- Is this still valid even if die Distribution for Python is no longer under the EULA for Intel Software Development Products?

- What about DLLs that are not listed within redist.txt (e.g. libifcoremd.dll)?

- Does this last sentence of the redist.txt apply für libifcoremd.dll: Also included as Redistributables are executables and shared and static libraries created by linking your program with libraries provided as part of the Materials for this product but which are not listed in this redist.txt file ("Libraries"); provided that you do not redistribute those libraries or any other portion of those libraries in any other way.

I would be happy to hear from anyone who has an opinion or further ideas regarding these topics.

I would be happy to read through much more about this. But with the Intel Simplified Software License having an emphasis on *simpliefied* I could not find much more. Is there any resource where I can dig into this any deeper?
 

Kind regards

[Robert_C_Intel]

It is ok to redistribute individual packages. It's ok to repackage them in a unified tar file. We pass through the licenses of the open source project, so look in pkgs/numpy*/info to understand the numpy license. Ignore the redist file, it looks like a leftover from the old EULA.

[Manuel_Z_]

Robert C. (Intel) wrote:

It is ok to redistribute individual packages. It's ok to repackage them in a unified tar file. We pass through the licenses of the open source project, so look in pkgs/numpy*/info to understand the numpy license. Ignore the redist file, it looks like a leftover from the old EULA.

Hi Robert, thanks for your advice.

Digging a bit deeper we found, that there might be the need to include some of the libraries (Windows DLLs) that can be found within $INTEL_PYTHON_DISTRIBUTION\Library\bin.

Do you see anything that we have to consider if we redistribute some these libraries? Can we simply assume all of them to be provided under Intel Simplified Software License?

Kind regards

[Robert_C_Intel]

We package everything in conda packages with a license from the owner. The installer unpacks the packages and puts the files where python requires them. To see the license for a file in the Library directory, look in the pkgs directory. One way you can do this is to create the minimal environment that your program needs. For example:

conda create -n myenv scipy

Activate the environment:

source activate myenv

Make sure it works and has everything you need and then do a

conda list

to see all the packages are used, then go to the pkgs directory to get the licenses for the packages.