libcurl/docs: expand on redirect following and secrets to other hosts

For CURLOPT_FOLLOWLOCATION and CURLOPT_UNRESTRICTED_AUTH

Ref: #14471
Closes #14472
This commit is contained in:
Daniel Stenberg 2024-08-09 13:58:19 +02:00
parent f0a551814b
commit 6f00a05e89
No known key found for this signature in database
GPG Key ID: 5CC908FDB71E12C2
2 changed files with 39 additions and 22 deletions

View File

@ -10,6 +10,7 @@ See-also:
- CURLOPT_POSTREDIR (3)
- CURLOPT_PROTOCOLS_STR (3)
- CURLOPT_REDIR_PROTOCOLS_STR (3)
- CURLOPT_UNRESTRICTED_AUTH (3)
Protocol:
- HTTP
Added-in: 7.1
@ -34,9 +35,9 @@ redirects that an HTTP server sends in a 30x response. The Location: header
can specify a relative or an absolute URL to follow.
libcurl issues another request for the new URL and follows subsequent new
Location: redirects all the way until no more such headers are returned or the
maximum limit is reached. CURLOPT_MAXREDIRS(3) is used to limit the
number of redirects libcurl follows.
`Location:` redirects all the way until no more such headers are returned or
the maximum limit is reached. CURLOPT_MAXREDIRS(3) is used to limit the number
of redirects libcurl follows.
libcurl restricts what protocols it automatically follow redirects to. The
accepted target protocols are set with CURLOPT_REDIR_PROTOCOLS_STR(3). By
@ -44,23 +45,32 @@ default libcurl allows HTTP, HTTPS, FTP and FTPS on redirects.
When following a redirect, the specific 30x response code also dictates which
request method libcurl uses in the subsequent request: For 301, 302 and 303
responses libcurl switches method from POST to GET unless
CURLOPT_POSTREDIR(3) instructs libcurl otherwise. All other redirect
response codes make libcurl use the same method again.
responses libcurl switches method from POST to GET unless CURLOPT_POSTREDIR(3)
instructs libcurl otherwise. All other redirect response codes make libcurl
use the same method again.
For users who think the existing location following is too naive, too simple
or just lacks features, it is easy to instead implement your own redirect
follow logic with the use of curl_easy_getinfo(3)'s
CURLINFO_REDIRECT_URL(3) option instead of using
CURLOPT_FOLLOWLOCATION(3).
follow logic with the use of curl_easy_getinfo(3)'s CURLINFO_REDIRECT_URL(3)
option instead of using CURLOPT_FOLLOWLOCATION(3).
By default, libcurl only sends `Authentication:` or explicitly set `Cookie:`
headers to the initial host given in the original URL, to avoid leaking
username + password to other sites. CURLOPT_UNRESTRICTED_AUTH(3) is provided
to change that behavior.
Due to the way HTTP works, almost any header can be made to contain data a
client may not want to pass on to other servers than the initially intended
host and for all other headers than the two mentioned above, there is no
protection from this happening when libcurl is told to follow redirects.
# NOTE
Since libcurl changes method or not based on the specific HTTP response code,
setting CURLOPT_CUSTOMREQUEST(3) while following redirects may change
what libcurl would otherwise do and if not that carefully may even make it
misbehave since CURLOPT_CUSTOMREQUEST(3) overrides the method libcurl
would otherwise select internally.
setting CURLOPT_CUSTOMREQUEST(3) while following redirects may change what
libcurl would otherwise do and if not that carefully may even make it
misbehave since CURLOPT_CUSTOMREQUEST(3) overrides the method libcurl would
otherwise select internally.
# DEFAULT

View File

@ -31,17 +31,19 @@ CURLcode curl_easy_setopt(CURL *handle, CURLOPT_UNRESTRICTED_AUTH,
# DESCRIPTION
Set the long *gohead* parameter to 1L to make libcurl continue to send
authentication (user+password) credentials when following locations, even when
hostname changed. This option is meaningful only when setting
CURLOPT_FOLLOWLOCATION(3).
authentication (user+password) credentials or explicitly set cookie headers
when following locations, even when the host changes. This option is
meaningful only when setting CURLOPT_FOLLOWLOCATION(3).
Further, when this option is not used or set to **0L**, libcurl does not
send custom nor internally generated Authentication: headers on requests done
to other hosts than the one used for the initial URL.
Further, when this option is not used or set to **0L**, libcurl does not send
custom nor internally generated `Authentication:` or `Cookie:` headers on
requests done to other hosts than the one used for the initial URL. Another
host means that one or more of hostname, protocol scheme or port number
changed.
By default, libcurl only sends credentials and Authentication headers to the
initial hostname as given in the original URL, to avoid leaking username +
password to other sites.
By default, libcurl only sends `Authentication:` or explicitly set `Cookie:`
headers to the initial host as given in the original URL, to avoid leaking
username + password to other sites.
This option should be used with caution: when curl follows redirects it
blindly fetches the next URL as instructed by the server. Setting
@ -49,6 +51,11 @@ CURLOPT_UNRESTRICTED_AUTH(3) to 1L makes curl trust the server and sends
possibly sensitive credentials to any host the server points to, possibly
again and again as the following hosts can keep redirecting to new hosts.
Due to the way HTTP works, almost any header can be made to contain data a
client may not want to pass on to other servers than the initially intended
host and for all other headers than the two mentioned above, there is no
protection from this happening when libcurl is told to follow redirects.
# DEFAULT
0