Closed yuvipanda closed 1 year ago
I've been very dissapointed with overall IPFS performance, and from the outside a lot of the focus seems to be somewhat filecoin related (and hence my brain just files it under 'oh, another grift'). Am no longer interested in this.
Corresponds with https://github.com/jupyterhub/repo2docker/issues/1096 in repo2docker, with the goal of being able to build and launch code from IPFS that can eventually also load data from IPFS, for a fully self contained demo.
Proposed change
IPFS is a content addressable global 'file system' that can share directories using an immutable, globaly unique content ID. It can be used to store code as well as data. There are experiments in the pydata / zarr ecosystem on using it for some datasets as well https://github.com/pangeo-forge/roadmap/issues/40.
I'd like us to add an IPFS content provider. Since IPFS content ID (CIDs) are immutable, we won't have to do much - just pass it on to repo2docker (once https://github.com/jupyterhub/repo2docker/issues/1096 is implemented)
Who would use this feature?
Ideally, this would eventually end up on mybinder.org and other binderhubs. IPFS can be a distributed alternative to storing code and content, vs something centralized like GitHub.