Prepare URL coerces to lower-case

[dmitrypolo]

requests/requests/models.py

Line 369 in 916d171

scheme, auth, host, port, path, query, fragment = parse_url(url)

When passing in a URL with an upper-case letter after a - character the host is coerced to lower case.

Python Version: 3.9.1
Requests Version: 2.26.0
OS: MacOS

Steps to reproduce

from requests.models import PreparedRequest
p = PreparedRequest()
url = "http://service-A/abc"

p.prepare_url(url, None)
print(p.url)
"http://service-a/abc"

Notice how the A has been changed to a. This behavior is not expected. I have also submitted this bug report as the behavior arises from the use of urllib3.

[sigmavirus24]

This is 100% correct. RFC 3986 describes the part of any URI as such:

• scheme (part of the URL preceding the first :)
• authority (part of the URL starting with // before the path; this includes username/password, hostname, and port)
• path (part of the URL starting with /)
• query (part of the URL starting with ?)
• fragment (part of the URL starting with #)

"http://service-A/abc" thusly gets parsed as scheme=http, authority=service-A, path=/abc. The host portion of the authority is to be treated case-insensitively. In other words, service-A is actually functionally equivalent to service-a.

Finally, spamming multiple projects is not appropriate, especially when you've done 0 research into how the standards define the behaviour here.

[TaquitoSlayer]

Imagine not realizing that we're getting to a point with Web where there are vastly more companies handling stuff like authorizations, header orders, and even case sensitivity per request, even in a production environment and dismissing a very real issue that is now becoming more common.

[sethmlarson]

@TaquitoSlayer Your sarcastic comment isn't appreciated. I suggest you read up on existing standards for the web before claiming you know how the web works:

• WHATWG URL which is the most modern standard for URLs calls for lowercasing the host section of URLs.

• RFC 3986 which is the standard that Requests and urllib3 primarily use as a reference for their URL parsers also calls for normalization of the host component via lowercasing (quote from Section 3.2.2):

  Although host is case-insensitive, producers and normalizers should use lowercase for registered names and hexadecimal addresses for the sake of uniformity, while only using uppercase letters for percent-encodings.

Anyone depending on any different behavior should expect their services to break completely as they're throwing 20+ years of standards out the window.