Replies: 2 comments 4 replies
-
You've understood perfectly - we have a separate service for handling uploads. It does very little from a "simple-repository" perspective, and it handles a bunch of different internal auth mechanisms. For this reason, we didn't include it in our release. There may be scope for this to change in the future, as we are gradually adding more "smarts" such as rejecting releases with invalid (or undesirable) metadata, as well as having some ownership mechanisms such as is seen in PyPI.
Yes, this totally works. In fact, I think this was the setup that we tested at EuroPython last year with the Pyodide repository. There, they have some controlled way of managing packages in their index, and build a custom JSON view of the index (they don't have a HTML simple equivalent, e.g. https://pyodide-pypi-api.s3.amazonaws.com/simple/numpy/ points to a JSON doc (with the right headers)). We were able to get the |
Beta Was this translation helpful? Give feedback.
-
As you use GitLab at CERN, do you use the GitLab PyPI endpoints as part of your internal structure to feed packages to the repository? |
Beta Was this translation helpful? Give feedback.
-
Looking over the code, I don't see endpoints for PUT requests, so I assume that's handled out of band. Thinking about our own setup (which would be in AWS), I can imagine running this purely as a proxy, with a "dumb" PEP-503 index built over s3 uploads for internal packages - just a lambda writing html index files - and rely on simple-repository's fancy ability to generate PEP-658 metadata on top?
How are you handling this yourselves (or have I completely misunderstood)
Beta Was this translation helpful? Give feedback.
All reactions