Repoint hardcoded bbop-sqlite S3 URLs to the semanticsql CDN before raw-bucket access is retired

[caufieldjh]

Repoint hardcoded bbop-sqlite S3 URLs to the SemanticSQL CDN

universalizer downloads pre-built OAK/SemanticSQL SQLite databases by hardcoding the raw S3 bucket URL. The maintainers of that bucket (Berkeley BOP / INCATools SemanticSQL) are retiring direct public access to the raw S3 bucket and moving all traffic behind a CDN. Because this URL is hardcoded (rather than resolved through OAK'''s sqlite:obo: selector or the semsql package), it will not auto-migrate and will break once raw-bucket access is removed.

Where this repo is affected

• universalizer/oak_utils.py — f"https://s3.amazonaws.com/bbop-sqlite/{db}.db"

What to change

• Old: https://s3.amazonaws.com/bbop-sqlite/{db}.db
• New: https://semanticsql.berkeleybop.io/{db}.db

Important

You must also send a non-default User-Agent. The CDN sits behind a Browser Integrity Check that returns HTTP 403 to default client User-Agents (Python's Python-urllib/3.x, R's download.file/httr, and bare curl/wget defaults). A host-only swap will start returning 403 — set an explicit, non-default User-Agent header on the request.

If you fetch with urllib/requests, add an explicit header, e.g. urllib.request.Request(url, headers={"User-Agent": "universalizer/1.0"}).

Why now

• Removal of raw-bucket public access is tracked in Eliminate S3 bucket access INCATools/semantic-sql#112.
• The object names/paths are identical — only the host changes (s3.amazonaws.com/bbop-sqlite/ or bbop-sqlite.s3.amazonaws.com/ → semanticsql.berkeleybop.io/).

Filed as part of the coordinated downstream migration tracked in INCATools/semantic-sql#115. If you've already migrated or this is a non-issue, feel free to close — thanks!