The file system semantics of DSX are too confusing today, and more importantly, they do not work with most of the Python machine learning ecosystem. FUSE allows user programs implementing filesystems to be "mounted" as a Linux filesystem. A quick search for object storage FUSE systems yields a number of projects that can expose an object storage API as a filesystem, and interestingly there is a page for mounting SoftLayer Object Storage that may (or may not) still be useful: https://knowledgelayer.softlayer.com/learning/how-mount-object-storage.
Is this similar to the databricks filesystem?
Attachments Open full size
Presumably this approach will remove the need for the “insert code” functionality for onject store? Some problems with “insert code” are that it is not supported on all languages/environments and it only provides functionality to read data and not for writing data which is a common need.
Attachments Open full size