Warn about urlsplit()'s `netloc` not being suitable for open redirect checking

>>> from urllib.parse import urlsplit
>>> urlsplit('//example.com')
SplitResult(scheme='', netloc='example.com', path='', query='', fragment='')
>>> urlsplit('///example.com')
SplitResult(scheme='', netloc='', path='/example.com', query='', fragment='')
>>> urlsplit('////example.com')
SplitResult(scheme='', netloc='', path='//example.com', query='', fragment='')

This greatly differs from how browsers interpret it: Location: ////example.com will redirect you to example.com, which makes checking for a non-empty netloc to avoid open redirects useless.

I tested using Firefox (which starts treating it as a netloc starting with 4 slashes). httpie on the other hand does not have this problem, and curl simply rejects such invalid redirects. So it is most likely a a case of browsers simply tolerating garbage instead of refusing it.

I think a warning in the docs that it is not suitable for checking against open redirects may be a nice thing to add...

Linked PRs

• gh-144448
• gh-151621

Inizia dalla documentazione di urllib.parse urlsplit() e esamina le PR collegate gh-144448 e gh-151621. Chiarisci la limitazione documentata per il controllo degli open redirect, quindi verifica che l'avviso rifletta accuratamente il comportamento del browser descritto nell'issue.