Design#

Goals#

The goal of the jupyterlab/rtc repository is to prototype a data model to support simultaneous distributed editing in Jupyter. Fundamentally, it’s about taking the existing concepts 1 provided by Jupyter Server and creating a near real time data model on top of them.

This would help provide a Google Docs like editing experience in Jupyter editors by allowing multiple simultaneous users to edit a document at once.

It would do this by moving much of the state to be shared on the server as well, which would also help address some other long-running issues with Jupyter, around preserving outputs even after you close your browser and dealing with race conditions around kernel management. This will happen in a number of different layers, added to the jupyterlab/rtc monorepo:

  1. Top-level layer developed in the Applications: JupyterLab, Spyder, Nteract…

  2. High-level layer, a.k.a Jupyter components that support editing all data in Jupyter server.

  3. Middle-level layer that expose friendly Real Time data access using React.js integration.

  4. Base-level layer that support CRDT implementations (Lumino and/or other CRDT libraries).

The definition of those layers is still in progress. You can discuss this on jupyterlab/rtc#61.

Non Goals#

There are also a number of other ways to improve collaboration in Jupyter which the jupyterlab/rtc repository is not meant to address, at least initially.

  • It is not meant to be able to synchronize notebooks or work between Jupyter servers running on multiple machines.

  • It doesn’t overlap with JupyterHub in that JupyterHub is about spawning multiple instances and this is about providing one instance for collaborative purposes.

  • It also isn’t about providing concepts of Teams, Projects, Groups, etc. These are definitely useful to have in a collaborative environment but for the initial work here we are assuming the same permission-less structure of a regular Jupyter server.

  • It’s also often useful to share what environment you are in with your notebook, to make it more reproducible. This isn’t about that kind of collaboration either.

1

Jupyter Server specifications.