This document assumes that you are familiar with HTML and general networking.
The increasing amount of applications moving to the web has made "HTTP Scripting" more frequently requested and wanted. To be able to automatically extract information from the web, to fake users, to post or upload data to web servers are all important tasks today.
Curl is a command line tool for doing all sorts of URL manipulations and
transfers, but this particular document focuses on how to use it when doing
HTTP requests for fun and profit. This documents assumes that you know how to
invoke curl --help
or curl --manual
to get basic information about it.
Curl is not written to do everything for you. It makes the requests, it gets the data, it sends data and it retrieves the information. You probably need to glue everything together using some kind of script language or repeated manual invokes.
HTTP is the protocol used to fetch data from web servers. It is a simple protocol that is built upon TCP/IP. The protocol also allows information to get sent to the server from the client using a few different methods, as is shown here.
HTTP is plain ASCII text lines being sent by the client to a server to request a particular action, and then the server replies a few text lines before the actual requested content is sent to the client.
The client, curl, sends an HTTP request. The request contains a method (like GET, POST, HEAD etc), a number of request headers and sometimes a request body. The HTTP server responds with a status line (indicating if things went well), response headers and most often also a response body. The "body" part is the plain data you requested, like the actual HTML or the image etc.
Using curl's option --verbose
(-v
as a short option) displays what kind of commands curl sends to the server,
as well as a few other informational texts.
--verbose
is the single most useful option when it comes to debug or even
understand the curl<->server interaction.
Sometimes even --verbose
is not enough. Then
--trace
and
--trace-ascii
offer even more details as they show everything curl sends and
receives. Use it like this:
curl --trace-ascii debugdump.txt http://www.example.com/
Many times you may wonder what exactly is taking all the time, or you just
want to know the amount of milliseconds between two points in a transfer. For
those, and other similar situations, the
--trace-time
option is
what you need. It prepends the time to each trace output line:
curl --trace-ascii d.txt --trace-time http://example.com/
When doing parallel transfers, it is relevant to see which transfer is doing
what. When response headers are received (and logged) you need to know which
transfer these are for.
--trace-ids
option is what
you need. It prepends the transfer and connection identifier to each trace
output line:
curl --trace-ascii d.txt --trace-ids http://example.com/
By default curl sends the response to stdout. You need to redirect it
somewhere to avoid that, most often that is done with -o
or -O
.
The Uniform Resource Locator format is how you specify the address of a particular resource on the Internet. You know these, you have seen URLs like https://curl.se or https://example.com a million times. RFC 3986 is the canonical spec. The formal name is not URL, it is URI.
The hostname is usually resolved using DNS or your /etc/hosts file to an IP address and that is what curl communicates with. Alternatively you specify the IP address directly in the URL instead of a name.
For development and other trying out situations, you can point to a different
IP address for a hostname than what would otherwise be used, by using curl's
--resolve
option:
curl --resolve www.example.org:80:127.0.0.1 http://www.example.org/
Each protocol curl supports operates on a default port number, be it over TCP or in some cases UDP. Normally you do not have to take that into consideration, but at times you run test servers on other ports or similar. Then you can specify the port number in the URL with a colon and a number immediately following the hostname. Like when doing HTTP to port 1234:
curl http://www.example.org:1234/
The port number you specify in the URL is the number that the server uses to offer its services. Sometimes you may use a proxy, and then you may need to specify that proxy's port number separately from what curl needs to connect to the server. Like when using an HTTP proxy on port 4321:
curl --proxy http://proxy.example.org:4321 http://remote.example.org/
Some services are setup to require HTTP authentication and then you need to provide name and password which is then transferred to the remote site in various ways depending on the exact authentication protocol used.
You can opt to either insert the user and password in the URL or you can provide them separately:
curl http://user:password@example.org/
or
curl -u user:password http://example.org/
You need to pay attention that this kind of HTTP authentication is not what is usually done and requested by user-oriented websites these days. They tend to use forms and cookies instead.
The path part is just sent off to the server to request that it sends back the associated response. The path is what is to the right side of the slash that follows the hostname and possibly port number.
The simplest and most common request/operation made using HTTP is to GET a URL. The URL could itself refer to a webpage, an image or a file. The client issues a GET request to the server and receives the document it asked for. If you issue the command line
curl https://curl.se
you get a webpage returned in your terminal window. The entire HTML document that that URL holds.
All HTTP replies contain a set of response headers that are normally hidden,
use curl's --include
(-i
)
option to display them as well as the rest of the document.
You can ask the remote server for ONLY the headers by using the
--head
(-I
) option which makes
curl issue a HEAD request. In some special cases servers deny the HEAD method
while others still work, which is a particular kind of annoyance.
The HEAD method is defined and made so that the server returns the headers
exactly the way it would do for a GET, but without a body. It means that you
may see a Content-Length:
in the response headers, but there must not be an
actual body in the HEAD response.
A single curl command line may involve one or many URLs. The most common case is probably to just use one, but you can specify any amount of URLs. Yes any. No limits. You then get requests repeated over and over for all the given URLs.
Example, send two GET requests:
curl http://url1.example.com http://url2.example.com
If you use --data
to POST to
the URL, using multiple URLs means that you send that same POST to all the
given URLs.
Example, send two POSTs:
curl --data name=curl http://url1.example.com http://url2.example.com
Sometimes you need to operate on several URLs in a single command line and do
different HTTP methods on each. For this, you might enjoy the
--next
option. It is basically a
separator that separates a bunch of options from the next. All the URLs
before --next
get the same method and get all the POST data merged into
one.
When curl reaches the --next
on the command line, it resets the method and
the POST data and allow a new set.
Perhaps this is best shown with a few examples. To send first a HEAD and then a GET:
curl -I http://example.com --next http://example.com
To first send a POST and then a GET:
curl -d score=10 http://example.com/post.cgi --next http://example.com/results.html
Forms are the general way a website can present an HTML page with fields for the user to enter data in, and then press some kind of 'OK' or 'Submit' button to get that data sent to the server. The server then typically uses the posted data to decide how to act. Like using the entered words to search in a database, or to add the info in a bug tracking system, display the entered address on a map or using the info as a login-prompt verifying that the user is allowed to see what it is about to see.
Of course there has to be some kind of program on the server end to receive the data you send. You cannot just invent something out of the air.
A GET-form uses the method GET, as specified in HTML like:
<form method="GET" action="junk.cgi">
<input type=text name="birthyear">
<input type=submit name=press value="OK">
</form>
In your favorite browser, this form appears with a text box to fill in and a
press-button labeled "OK". If you fill in '1905' and press the OK button,
your browser then creates a new URL to get for you. The URL gets
junk.cgi?birthyear=1905&press=OK
appended to the path part of the previous
URL.
If the original form was seen on the page www.example.com/when/birth.html
,
the second page you get becomes
www.example.com/when/junk.cgi?birthyear=1905&press=OK
.
Most search engines work this way.
To make curl do the GET form post for you, just enter the expected created URL:
curl "http://www.example.com/when/junk.cgi?birthyear=1905&press=OK"
The GET method makes all input field names get displayed in the URL field of your browser. That is generally a good thing when you want to be able to bookmark that page with your given data, but it is an obvious disadvantage if you entered secret information in one of the fields or if there are a large amount of fields creating a long and unreadable URL.
The HTTP protocol then offers the POST method. This way the client sends the data separated from the URL and thus you do not see any of it in the URL address field.
The form would look similar to the previous one:
<form method="POST" action="junk.cgi">
<input type=text name="birthyear">
<input type=submit name=press value=" OK ">
</form>
And to use curl to post this form with the same data filled in as before, we could do it like:
curl --data "birthyear=1905&press=%20OK%20" http://www.example.com/when/junk.cgi
This kind of POST uses the Content-Type application/x-www-form-urlencoded
and is the most widely used POST kind.
The data you send to the server MUST already be properly encoded, curl does
not do that for you. For example, if you want the data to contain a space,
you need to replace that space with %20
, etc. Failing to comply with this
most likely causes your data to be received wrongly and messed up.
Recent curl versions can in fact url-encode POST data for you, like this:
curl --data-urlencode "name=I am Daniel" http://www.example.com
If you repeat --data
several times on the command line, curl concatenates
all the given data pieces - and put a &
symbol between each data segment.
Back in late 1995 they defined an additional way to post data over HTTP. It is documented in the RFC 1867, why this method sometimes is referred to as RFC 1867-posting.
This method is mainly designed to better support file uploads. A form that allows a user to upload a file could be written like this in HTML:
<form method="POST" enctype='multipart/form-data' action="upload.cgi">
<input name=upload type=file>
<input type=submit name=press value="OK">
</form>
This clearly shows that the Content-Type about to be sent is
multipart/form-data
.
To post to a form like this with curl, you enter a command line like:
curl --form upload=@localfilename --form press=OK [URL]
A common way for HTML based applications to pass state information between pages is to add hidden fields to the forms. Hidden fields are already filled in, they are not displayed to the user and they get passed along just as all the other fields.
A similar example form with one visible field, one hidden field and one submit button could look like:
<form method="POST" action="foobar.cgi">
<input type=text name="birthyear">
<input type=hidden name="person" value="daniel">
<input type=submit name="press" value="OK">
</form>
To POST this with curl, you do not have to think about if the fields are hidden or not. To curl they are all the same:
curl --data "birthyear=1905&press=OK&person=daniel" [URL]
When you are about to fill in a form and send it to a server by using curl instead of a browser, you are of course interested in sending a POST exactly the way your browser does.
An easy way to get to see this, is to save the HTML page with the form on your local disk, modify the 'method' to a GET, and press the submit button (you could also change the action URL if you want to).
You then clearly see the data get appended to the URL, separated with a
?
-letter as GET forms are supposed to.
Perhaps the best way to upload data to an HTTP server is to use PUT. Then again, this of course requires that someone put a program or script on the server end that knows how to receive an HTTP PUT stream.
Put a file to an HTTP server with curl:
curl --upload-file uploadfile http://www.example.com/receive.cgi
HTTP Authentication is the ability to tell the server your username and password so that it can verify that you are allowed to do the request you are doing. The Basic authentication used in HTTP (which is the type curl uses by default) is plain text based, which means it sends username and password only slightly obfuscated, but still fully readable by anyone that sniffs on the network between you and the remote server.
To tell curl to use a user and password for authentication:
curl --user name:password http://www.example.com
The site might require a different authentication method (check the headers
returned by the server), and then
--ntlm
,
--digest
,
--negotiate
or even
--anyauth
might be
options that suit you.
Sometimes your HTTP access is only available through the use of an HTTP proxy. This seems to be especially common at various companies. An HTTP proxy may require its own user and password to allow the client to get through to the Internet. To specify those with curl, run something like:
curl --proxy-user proxyuser:proxypassword curl.se
If your proxy requires the authentication to be done using the NTLM method,
use --proxy-ntlm
, if
it requires Digest use
--proxy-digest
.
If you use any one of these user+password options but leave out the password part, curl prompts for the password interactively.
Do note that when a program is run, its parameters might be possible to see when listing the running processes of the system. Thus, other users may be able to watch your passwords if you pass them as plain command line options. There are ways to circumvent this.
It is worth noting that while this is how HTTP Authentication works, many websites do not use this concept when they provide logins etc. See the Web Login chapter further below for more details on that.
An HTTP request may include a 'referer' field (yes it is misspelled), which can be used to tell from which URL the client got to this particular resource. Some programs/scripts check the referer field of requests to verify that this was not arriving from an external site or an unknown page. While this is a stupid way to check something so easily forged, many scripts still do it. Using curl, you can put anything you want in the referer-field and thus more easily be able to fool the server into serving your request.
Use curl to set the referer field with:
curl --referer http://www.example.come http://www.example.com
Similar to the referer field, all HTTP requests may set the User-Agent field. It names what user agent (client) that is being used. Many applications use this information to decide how to display pages. Silly web programmers try to make different pages for users of different browsers to make them look the best possible for their particular browsers. They usually also do different kinds of JavaScript etc.
At times, you may learn that getting a page with curl does not return the same page that you see when getting the page with your browser. Then you know it is time to set the User Agent field to fool the server into thinking you are one of those browsers.
By default, curl uses curl/VERSION, such as User-Agent: curl/8.11.0.
To make curl look like Internet Explorer 5 on a Windows 2000 box:
curl --user-agent "Mozilla/4.0 (compatible; MSIE 5.01; Windows NT 5.0)" [URL]
Or why not look like you are using Netscape 4.73 on an old Linux box:
curl --user-agent "Mozilla/4.73 [en] (X11; U; Linux 2.2.15 i686)" [URL]
When a resource is requested from a server, the reply from the server may
include a hint about where the browser should go next to find this page, or a
new page keeping newly generated output. The header that tells the browser to
redirect is Location:
.
Curl does not follow Location:
headers by default, but simply displays such
pages in the same manner it displays all HTTP replies. It does however
feature an option that makes it attempt to follow the Location:
pointers.
To tell curl to follow a Location:
curl --location http://www.example.com
If you use curl to POST to a site that immediately redirects you to another
page, you can safely use --location
(-L
) and --data
/--form
together. Curl only uses POST in the first
request, and then revert to GET in the following operations.
Browsers typically support at least two other ways of redirects that curl does not: first the html may contain a meta refresh tag that asks the browser to load a specific URL after a set number of seconds, or it may use JavaScript to do it.
The way the web browsers do "client side state control" is by using cookies. Cookies are just names with associated contents. The cookies are sent to the client by the server. The server tells the client for what path and hostname it wants the cookie sent back, and it also sends an expiration date and a few more properties.
When a client communicates with a server with a name and path as previously specified in a received cookie, the client sends back the cookies and their contents to the server, unless of course they are expired.
Many applications and servers use this method to connect a series of requests into a single logical session. To be able to use curl in such occasions, we must be able to record and send back cookies the way the web application expects them. The same way browsers deal with them.
The simplest way to send a few cookies to the server when getting a page with curl is to add them on the command line like:
curl --cookie "name=Daniel" http://www.example.com
Cookies are sent as common HTTP headers. This is practical as it allows curl
to record cookies simply by recording headers. Record cookies with curl by
using the --dump-header
(-D
)
option like:
curl --dump-header headers_and_cookies http://www.example.com
(Take note that the
--cookie-jar
option described
below is a better way to store cookies.)
Curl has a full blown cookie parsing engine built-in that comes in use if you want to reconnect to a server and use cookies that were stored from a previous connection (or hand-crafted manually to fool the server into believing you had a previous connection). To use previously stored cookies, you run curl like:
curl --cookie stored_cookies_in_file http://www.example.com
Curl's "cookie engine" gets enabled when you use the
--cookie
option. If you only
want curl to understand received cookies, use --cookie
with a file that
does not exist. Example, if you want to let curl understand cookies from a
page and follow a location (and thus possibly send back cookies it received),
you can invoke it like:
curl --cookie nada --location http://www.example.com
Curl has the ability to read and write cookie files that use the same file
format that Netscape and Mozilla once used. It is a convenient way to share
cookies between scripts or invokes. The --cookie
(-b
) switch
automatically detects if a given file is such a cookie file and parses it,
and by using the --cookie-jar
(-c
) option you make curl write a new
cookie file at the end of an operation:
curl --cookie cookies.txt --cookie-jar newcookies.txt \
http://www.example.com
There are a few ways to do secure HTTP transfers. By far the most common protocol for doing this is what is generally known as HTTPS, HTTP over SSL. SSL encrypts all the data that is sent and received over the network and thus makes it harder for attackers to spy on sensitive information.
SSL (or TLS as the current version of the standard is called) offers a set of advanced features to do secure transfers over HTTP.
Curl supports encrypted fetches when built to use a TLS library and it can be
built to use one out of a fairly large set of libraries - curl -V
shows
which one your curl was built to use (if any). To get a page from an HTTPS
server, simply run curl like:
curl https://secure.example.com
In the HTTPS world, you use certificates to validate that you are the one you claim to be, as an addition to normal passwords. Curl supports client- side certificates. All certificates are locked with a passphrase, which you need to enter before the certificate can be used by curl. The passphrase can be specified on the command line or if not, entered interactively when curl queries for it. Use a certificate with curl on an HTTPS server like:
curl --cert mycert.pem https://secure.example.com
curl also tries to verify that the server is who it claims to be, by
verifying the server's certificate against a locally stored CA cert bundle.
Failing the verification causes curl to deny the connection. You must then
use --insecure
(-k
) in case you
want to tell curl to ignore that the server cannot be verified.
More about server certificate verification and ca cert bundles can be read in
the SSLCERTS
document.
At times you may end up with your own CA cert store and then you can tell curl to use that to verify the server's certificate:
curl --cacert ca-bundle.pem https://example.com/
Doing fancy stuff, you may need to add or change elements of a single curl request.
For example, you can change the POST method to PROPFIND
and send the data
as Content-Type: text/xml
(instead of the default Content-Type
) like
this:
curl --data "<xml>" --header "Content-Type: text/xml" \
--request PROPFIND example.com
You can delete a default header by providing one without content. Like you
can ruin the request by chopping off the Host:
header:
curl --header "Host:" http://www.example.com
You can add headers the same way. Your server may want a Destination:
header, and you can add it:
curl --header "Destination: http://nowhere" http://example.com
It should be noted that curl selects which methods to use on its own
depending on what action to ask for. -d
makes a POST, -I
makes a HEAD and
so on. If you use the --request
/
-X
option you can change the method keyword curl selects, but you do not
modify curl's behavior. This means that if you for example use -d "data" to
do a POST, you can modify the method to a PROPFIND
with -X
and curl still
thinks it sends a POST. You can change the normal GET to a POST method by
simply adding -X POST
in a command line like:
curl -X POST http://example.org/
curl however still acts as if it sent a GET so it does not send any request body etc.
While not strictly just HTTP related, it still causes a lot of people problems so here's the executive run-down of how the vast majority of all login forms work and how to login to them using curl.
It can also be noted that to do this properly in an automated fashion, you most certainly need to script things and do multiple curl invokes etc.
First, servers mostly use cookies to track the logged-in status of the client, so you need to capture the cookies you receive in the responses. Then, many sites also set a special cookie on the login page (to make sure you got there through their login page) so you should make a habit of first getting the login-form page to capture the cookies set there.
Some web-based login systems feature various amounts of JavaScript, and sometimes they use such code to set or modify cookie contents. Possibly they do that to prevent programmed logins, like this manual describes how to... Anyway, if reading the code is not enough to let you repeat the behavior manually, capturing the HTTP requests done by your browsers and analyzing the sent cookies is usually a working method to work out how to shortcut the JavaScript need.
In the actual <form>
tag for the login, lots of sites fill-in
random/session or otherwise secretly generated hidden tags and you may need
to first capture the HTML code for the login form and extract all the hidden
fields to be able to do a proper login POST. Remember that the contents need
to be URL encoded when sent in a normal POST.
Many times when you run curl on a site, you notice that the site does not seem to respond the same way to your curl requests as it does to your browser's.
Then you need to start making your curl requests more similar to your browser's requests:
Use the --trace-ascii
option to store fully detailed logs of the requests
for easier analyzing and better understanding
Make sure you check for and use cookies when needed (both reading with
--cookie
and writing with --cookie-jar
)
Set user-agent (with -A
) to
one like a recent popular browser does
Set referer (with -E
) like
it is set by the browser
If you use POST, make sure you send all the fields and in the same order as the browser does it.
A good helper to make sure you do this right, is the web browsers' developers tools that let you view all headers you send and receive (even when using HTTPS).
A more raw approach is to capture the HTTP traffic on the network with tools
such as Wireshark or tcpdump and check what headers that were sent and
received by the browser. (HTTPS forces you to use SSLKEYLOGFILE
to do
that.)