QA rating without downloading files

I would like to suggest that the extension be fundamentally redesigned so that it is not necessary to download complete files.

It is not ideal to download the complete files. Some files are several gigabyte large. For other resources, a lot of computing time is required in the source system.

Therefore, I suggest different levels to determine the file format:

1. Trust the format specification made by the user.
2. Do a HTTP HEAD request and trust the webserver's answer.
3. Download a few bytes of the resource and use *file magic numbers*.
4. Download the complete file and do the analysis.

The issue names no files, tests, or entry points, so the first step is to identify where resource downloading and format detection are handled. Done would mean supporting an agreed format-detection level that avoids downloading complete files where possible.

Written by the indexing model from the issue text.