They don't retry on errors but they know it is an error. Eg. imagine a shell script using curl or wget and trying multiple URLs as a health check (eg. on different round-robin IPs). Without these "generic" HTTP tools knowing that this is a "failure", you would need to implement custom parsing for any case like this instead of relying on the defined "error" and "success" behaviour.
The same holds true if you are using any programming library: there is a plethora of handlers for HTTP errors.
As for redirection, a common example is offering downloads through S3 using pre-signed URLs (you share a URL with your own domain, but after auth redirect to a pre-signed S3 URL for direct download or upload).
The same holds true if you are using any programming library: there is a plethora of handlers for HTTP errors.
As for redirection, a common example is offering downloads through S3 using pre-signed URLs (you share a URL with your own domain, but after auth redirect to a pre-signed S3 URL for direct download or upload).