Inconsistent filenames when url download redirects

[philwinder]

When I run this:

bacalhau docker run --input-urls=https://github.com/ultralytics/yolov5/releases/download/v6.1/yolov5s.pt ubuntu -- ls -lah /inputs

The result is:

total 15M
drwxr-xr-x 2 root root 4.0K Nov 24 15:12 .
drwxr-xr-x 1 root root 4.0K Nov 24 15:12 ..
-rw-r--r-- 1 root root  15M Nov 24 15:12 76813c2d-b52b-47af-95fb-e92c1b0b2783

When curling that URL you'll see that it redirects to another with a (random?) UUID for the name. Something that looks like this:

https://objects.githubusercontent.com/github-production-release-asset-2e65be/264818686/14327886-3839-4fa5-96c3-d52cfa73cdc5?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20221124%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20221124T151717Z&X-Amz-Expires=300&X-Amz-Signature=c3a324c90a4f6ac3e54d6cdeefbfddc70ccd0472667cfb9652a3f0e15ee37fa1&X-Amz-SignedHeaders=host&actor_id=0&key_id=0&repo_id=264818686&response-content-disposition=attachment%3B%20filename%3Dyolov5s.pt&response-content-type=application%2Foctet-stream

As a user I expect the file name to match the base name of the original URL, not the redirected one.

An alternative fix would be to allow volume mapping to rename foo:bar

Source: https://filecoinproject.slack.com/archives/C02RLM3JHUY/p1669302839546539

[aronchick]

The problem is the opposite is true too - https://picsum.photos/200/200/ - redirects to get the final file name.

This: https://superuser.com/questions/301044/how-to-wget-a-file-with-correct-name-when-redirected <- suggests we should look at content-disposition. CURL doesn't work unless you give it an -o to name the output. wget uses the original file name.

I've seen many sites out there with things like this - http://www.vim.org/scripts/download_script.php?src_id=9750 and i think that naming things download_script.php would not be expected either.