[Bug]: normalize_url lowercases the userinfo in a URL, so credentials in a link are rewritten

Ouverte
#2,529 0 commentaires 0 réactions 0 personnes assignées Voir sur GitHub

Les mainteneurs répondent en général sous 1 jour

@Steve-20-ai y travaille déjà.

Depuis le 10/11/2026.

  • #2530 par @Steve-20-ai — ouverte

Évaluation

Difficulté
2/5
Temps estimé
1-3 heures
Accessibilité débutants
20/100
Type d'issue
Bug
Clarté
Clairement spécifiée
Activité
À l'abandon
Stack technique
python
Domaine
backend

Piste de recherche

Read the three normalizers named in the issue: normalize_url, normalize_url_for_deep_crawl, and efficient_normalize_url_for_deep_crawl. Check how each handles parsed.netloc and add regression coverage for preserved userinfo casing and lowercase hostnames. The issue says the author has a patch and six tests ready, and linked PR #2530 is open.

Rédigé par le modèle d'indexation à partir du texte de l'issue.

Description

Describe the bug

normalize_url, normalize_url_for_deep_crawl and efficient_normalize_url_for_deep_crawl all lowercase the whole netloc:

netloc = parsed.netloc.lower()

netloc also carries the user:password@ part of a URL. A hostname is case-insensitive, so lowering it is right; credentials are not, so the same call silently rewrites them. The comment on the efficient normalizer even says what was intended — # 1. Lowercase hostname.

Steps to Reproduce

from crawl4ai.utils import normalize_url, normalize_url_for_deep_crawl

base = "https://example.com/dir/page"
print(normalize_url("https://User:[email protected]/path", base))
print(normalize_url_for_deep_crawl("https://User:[email protected]/path", base))

Actual:

https://user:[email protected]/path
https://user:[email protected]/path

Expected:

https://User:[email protected]/path
https://User:[email protected]/path

Why it matters

quick_extract_links resolves every <a href> through normalize_url_for_deep_crawl, so any link carrying credentials in its URL is rewritten before the crawl follows it. A page that links to https://user:SecRet@host/ is then fetched with user:secret, and the server answers 401 — or, worse, authenticates as a different account on systems where the userinfo is an identifier rather than a password. The same happens to URLs assembled by hand and passed to the deep-crawl path.

Environment

  • crawl4ai main at cae74b2
  • Python 3.13, Windows
  • No browser involved: the three functions are pure string handling

Proposed patch

Lowercase only the host, keeping whatever precedes the @ as written:

def _lowercase_host(netloc: str) -> str:
    userinfo, separator, hostport = netloc.rpartition('@')
    return f"{userinfo}{separator}{hostport.lower()}"

rpartition keeps a %40-encoded @ inside the password intact, since urlparse leaves it encoded in the netloc. Used at the three call sites, the diff is one line each.

Behaviour for URLs without userinfo is unchanged, including the port: Example.COM:8080 still becomes example.com:8080.

I have the patch and six regression tests ready (userinfo casing, host-only casing, userinfo with port, encoded @ in the password, and the two deep-crawl normalizers) and will open the PR against this issue.

Langage dominant
Python
Étoiles
87.2k
Forks
9.1k
Merge moyen
1 j 6 h
PR mergées (30 j)
25

Préparer son environnement

Par où commencer

  1. Lisez l'issue en entier, puis le guide de contribution du projet.
  2. Signalez en commentaire que vous la prenez — cela évite que deux personnes fassent le même travail.
  3. Forkez le dépôt et travaillez sur une branche.
  4. Ouvrez une pull request qui référence le numéro de l'issue.

Autres issues de unclecode/crawl4ai

Toutes les issues de unclecode/crawl4ai

Issues similaires

Plus d'issues Python

Recevez les nouvelles issues par e-mail

Un résumé court des issues GitHub adaptées aux débutants.