drivendataorg/cloudpathlib

az:// URL scheme isn't really a URL because it is not unique

Open

#157 opened on Aug 28, 2021

 (3 comments) (0 reactions) (0 assignees)Python (87 forks)auto 404
Azurehelp wanted

Repository metrics

Stars
 (624 stars)
PR merge metrics
 (PR metrics pending)

Description

The existing 'az://' pseudo-URL scheme violates the principal of URLs in that it is not universal and unique.

The 's3://' bucket syntax is valid because S3 buckets are universally unique. You may need account credentials to access an S3 bucket but you don't need additional information to locate it.

With Azure blob storage you need the storage account as well. Without that you have no way to locate the blobs even if they are publically accessible.

A better scheme would be something like:

az://<storage-account>/<container>/...

which would turn into:

https://<storage-account>.blob.core.windows.net/<container>/...

You could allow the optional addition of .blob or .dfs, so one could write:

az://<storage-account>.dfs/<container>/...

Contributor guide