Kevin Bates

> So now, I wait and think about you. You're facing this as well, huh? Sucks to be here, I know. Hi @arjunv - we appreciate the thoughts and, yes,...

Hi @mgorny. Do you happen to run with [`JUPYTER_PATH`](https://github.com/jupyter/jupyter_core/blob/ac0898c3ecc5e49b2e61fed1adb9c8391606adff/jupyter_core/paths.py#L155-L156) defined? This will side-effect that particular test. FWIW, I can only reproduce the issue if I set `JUPYTER_PATH`. @jasongrout - would...

Sorry about this Michał. I'm hoping Jason can take a look at this. In the meantime, could you please post the results of: ``` jupyter --paths --debug ```

Hmm - I suspect your [ENV and SYSTEM paths are the same](https://github.com/jupyter/jupyter_core/blob/master/jupyter_core/paths.py#L164) in that your python is installed is `/usr/share/jupyter` - which will also side-effect the test.

What does this python snippet produce? ```python import sys sys.prefix ``` If either `/usr/local/share/jupyter` or `/usr/share/jupyter`, then that's the issue.

Yes, I would advise 3.5 clients to only downgrade to `6.1.12` as there is [considerable functionality (used by downstream applications) present between `6.0.0` and `6.1.12`](https://github.com/jupyter/jupyter_client/blob/master/docs/changelog.rst). Release `6.1.13` has been yanked...

Hi @JohanMabille - this is exactly what we do in Jupyter Enterprise Gateway - let the kernel (actually a kernel _wrapper_) determine its ports and return the connection info back...

@SylvainCorlay - yes, this is identical to what we already do in EG, so I'd like to make sure things are done in a compatible way and I'd love to...

> We can of course pick a random port and listen to another one for each kernel, but why do it ? The only reasonable reason would be to use...

Wouldn't kernels need to tolerate both approaches and, if passed a handshake-port call `write_connectionfile()` (essentially), otherwise use the connection file? Or, I want to say Matthias mentioned this above, embed...