# Meeting with Satra

## Authentication -> Authorization

We want to:

1. Read data from some source
- This can be done in many ways: custom api call, mounted list etc
3. How to decide authorization dynamically
- Configuring allowed_users? https://github.com/jupyterhub/jupyterhub/blob/11f00dbbe794191d328c6c2a44a9e85f7eb6ede5/jupyterhub/auth.py#L270-L304
- eading of ACL on startup is not great: https://github.com/neurohackademy/nh2020-jupyterhub/blob/32ab7bc2d24717a8d732da7e442c2ab6a1bc2daa/chart/values.yaml#L44-L72

## Profile list

## Enable accessing services launched on the pod

## Enable MATLAB kernel

## User based culling

To let the jupyterhub-idle-culler make decisions based on some metadata about
the server.
- Pod labels?
- This is not great, because jupyterhub-idle-culler is agnostic to k8s.
- Spawner state?
- Does the jupyterhub-idle-culler get information about the spawner that in turn may hold state?

### Mybinder.org culling mechanism

I think this is the configuration that makes mybinder.org cull user pods after a certain time and such
https://github.com/jupyterhub/mybinder.org-deploy/blob/master/mybinder/values.yaml#L227-L233

## Quotas (potentially via NFS)

- Is it easier to acquire quotas via providing storage via S3 buckets?
- How to enforce quotas via NFS?
- Self hosted NFS + XFS disk works
- How to acquire a XFS disk cheaply?
- Is it possible expose XFS via EFS, like a translation of the File System type.

---

Potential features:
- storage quotas (my use case: EFS)
- group based resources using yaml files (a la neurohackademy)
-