Delicious Bookmark this on Delicious Share on Facebook SlashdotSlashdot It! Digg! Digg

PHP : Function Reference : Network Functions : setcookie


Send a cookie (PHP 4, PHP 5)
bool setcookie ( string name [, string value [, int expire [, string path [, string domain [, bool secure [, bool httponly]]]]]] )

Example 1609. setcookie() send example

= 'something from somewhere';

setcookie("TestCookie", $value);
setcookie("TestCookie", $value, time()+3600);  /* expire in 1 hour */
setcookie("TestCookie", $value, time()+3600, "/~rasmus/", "", 1);

Example 1610. setcookie() delete example

When deleting a cookie you should assure that the expiration date is in the past, to trigger the removal mechanism in your browser. Examples follow how to delete cookies sent in previous example:

// set the expiration date to one hour ago
setcookie ("TestCookie", "", time() - 3600);
setcookie ("TestCookie", "", time() - 3600, "/~rasmus/", "", 1);

Example 1611. setcookie() and arrays

You may also set array cookies by using array notation in the cookie name. This has the effect of setting as many cookies as you have array elements, but when the cookie is received by your script, the values are all placed in an array with the cookie's name:

// set the cookies
setcookie("cookie[three]", "cookiethree");
setcookie("cookie[two]", "cookietwo");
setcookie("cookie[one]", "cookieone");

// after the page reloads, print them out
if (isset($_COOKIE['cookie'])) {
   foreach (
$_COOKIE['cookie'] as $name => $value) {
"$name : $value <br />\n";

The above example will output:

three : cookiethree
two : cookietwo
one : cookieone

Related Examples ( Source code ) » setcookie

Code Examples / Notes » setcookie


[Editor's note: ... or simply use sessions. You can't be sure that the visitor will use the same IP the next visit. Not even on the next request (thanks to proxy servers)]
Note on setting cookies allowing access to sites:
If you are not using something "personal" from the computer that you are sending the cookie too watch out.  Via javascript it is possible to steal cookies from other users.  Thus allowing the stealer to login to your site as another user that might not have access otherwise.  Try to add something like the user's ip in the cookie and allowing access from that ip only with the stored cookie data.


You can use cookies to prevent a browser refresh repeating some action from a form post... (providing the client is cookie enabled!)
//Flag up repeat actions (like credit card transaction, etc)
if(count($_POST)>0) {
   $lastpost= isset($_COOKIE['lastpost']) ? $_COOKIE['lastpost'] : '';
   if($lastpost!=md5(serialize($_POST))) {
       setcookie('lastpost', md5(serialize($_POST)));
   } else {
//At this point, if $_POST['_REPEATED']==1, then  the user
//has hit the refresh button; so don't do any actions that you don't
//want to repeat!
Hope that helps :)


You can be sure about the cookie files contents weren't changed.
$Seperator = '--';
$uniqueID = 'Ju?hG&F0yh9?=/6*GVfd-d8u6f86hp';
$Data = 'Ahmet '.md5('123456789');
setcookie('VerifyUser', $Data.$Seperator.md5($Data.$uniqueID));
if ($_COOKIE) {
  $Cut = explode($Seperator, $_COOKIE['VerifyUser']);
  if (md5($Cut[0].$uniqueID) === $Cut[1]) {
      $_COOKIE['VerifyUser'] = $Cut[0];
  } else {
      die('Cookie data is invalid!!!');
echo $_COOKIE['VerifyUser'];
Create a unique id for your site and create a hash with md5($Data.$uniqueID). Attacker can understant that it must be re-hash after change cookie content.
But doesn't. Because cannot guess your unique id. Seperate your hash and data with seperator and send that cookie. Control that hash of returned value and your unique id's is same returned hash. Otherwise you have to stop attack. Sorry for my poor english!


You always can "emulate" a php setcookies invoking javascript.
echo "<script type='text/javascript'>/n";
echo "document.cookie='example=cookie'; /n";
echo "</script>/n";
Using javascript to set a cookie allow you to change the cookie anytime (not only prior the body) running javascript directly in the client side.
The disadvantage is the clients must support javascript.

j├Ârg aldinger

When using your cookies on a webserver that is not on the standard port 80, you should NOT include the :[port] in the "Cookie domain" parameter, since this would not be recognized correctly.
I had the issue working on a project that runs on multiple servers (development, production, etc.). One of the servers is running on a different port (together with other websites that run on the same server but on different ports).


When setting a top level domain ex "" you must add the secure arg so it should look like
setcookie("TestCookie", $value, time()+3600, "/", "", 0);
ignoring the secure arg makes IE ignores the cookie
to get the top level domain use
$myDomain = ereg_replace('^[^\.]*\.([^\.]*)\.(.*)$', '\1.\2',$_SERVER['HTTP_HOST']);
to avoid localhost switch use
$phpCkDmn = $_SERVER['HTTP_HOST'] != "localhost" ? $myDomain : false;


using time()-1 to delete (expire) a cookie only works if the client's clock is set exact. my testing showed some weird results with my clock set 1 second or so behind the server. 1 day or even just a few seconds ahead of the server's time and the cookie doesn't expire when it's suposed to.
my test:
echo $k;
setting the expire time to 0 makes it a browser session cookie, lasting forever, until the browser is closed. but setting the expire time to 1 is the lowest timestamp possible and is most likely to expire the cookie without any problems.
my fix:
echo $k;
this is my theory. I'm not sure why no one else has thought of this problem or solution, and I'm still testing, but please email me your questions or comments.


This limitation is mentiones in the <a href=””> Netscape HTTP Cookies Preliminary Specification </a> above but I thought it might save someone some time if it was posted here. (I wasted almost a day about it, Serves me right for not RTFMing ...)
There are limitations on the number of cookies that a client can store at any one time. This is a specification of the minimum number of cookies that a client should be prepared to receive and store.
   * 300 total cookies
   * 4 kilobytes per cookie, where the name and the OPAQUE_STRING combine to form the 4 kilobyte limit.
   * 20 cookies per server or domain. (note that completely specified hosts and domains are treated as separate entities and have a 20 cookie limitation for each, not combined)
Servers should not expect clients to be able to exceed these limits. When the 300 cookie limit or the 20 cookie per server limit is exceeded, clients should delete the least recently used cookie. When a cookie larger than 4 kilobytes is encountered the cookie should be trimmed to fit, but the name should remain intact as long as it is less than 4 kilobytes.


This code sets cookie with Max-Age.
See to:
function set_cookie($Name, $Value = '', $MaxAge = 0, $Path = '', $Domain = '', $Secure = false, $HTTPOnly = false) {
 header('Set-Cookie: ' . rawurlencode($Name) . '=' . rawurlencode($Value)
                       . (empty($MaxAge) ? '' : '; Max-Age=' . $MaxAge)
                       . (empty($Path)   ? '' : '; path=' . $Path)
                       . (empty($Domain) ? '' : '; domain=' . $Domain)
                       . (!$Secure       ? '' : '; secure')
                       . (!$HTTPOnly     ? '' : '; HttpOnly'), false);
# examples:
set_cookie("TestCookie", $value, 3600);  /* expire in 1 hour */
set_cookie("TestCookie", $value, 3600, "/~rasmus/", "", 1);


There are several characters that dont work correctly if part of the cookie name, I in particular ran into problems using = as part of the cookie name ie/
setcookie('dE4fR=', $foo);
this was as a result of base64_encode() ing a variable to use as the name. The cookie was being set fine but the = is stored as a little square character in the cookie which prevent proper retreival.
The solution for me was to use md5() instead of base64_encode() but any other method which avoids the = sign should also solve the problem


The timeout is tricky as far as I read it depends on the time synchronization between the server and the client. If the GMT time is different (for example your computer is 5 min forward from the server time, don't know about the daylight saving problem) then the cookie timeout if affected by this.
Since I didn't very this information don't take it as factual but as a possible reason for problem (like the one below about IE).
The second conclusion is to use big values. For small values use something else (sessions, database...).


The following code snippet combines abdullah's and Charles Martin's examples into a powerful combination function (and fixes at least one bug in the process):
 function set_cookie_fix_domain($Name, $Value = '', $Expires = 0, $Path = '', $Domain = '', $Secure = false, $HTTPOnly = false)
   if (!empty($Domain))
     // Fix the domain to accept domains with and without 'www.'.
     if (strtolower(substr($Domain, 0, 4)) == 'www.')  $Domain = substr($Domain, 4);
     $Domain = '.' . $Domain;
     // Remove port information.
     $Port = strpos($Domain, ':');
     if ($Port !== false)  $Domain = substr($Domain, 0, $Port);
   header('Set-Cookie: ' . rawurlencode($Name) . '=' . rawurlencode($Value)
                         . (empty($Expires) ? '' : '; expires=' . gmdate('D, d-M-Y H:i:s', $Expires) . ' GMT')
                         . (empty($Path) ? '' : '; path=' . $Path)
                         . (empty($Domain) ? '' : '; domain=' . $Domain)
                         . (!$Secure ? '' : '; secure')
                         . (!$HTTPOnly ? '' : '; HttpOnly'), false);
Basically, if the domain parameter is supplied, it is converted to support a broader range of domains.  This behavior may or may not be desireable (e.g. could be a security problem depending on the server) but it makes cookie handling oh-so-much-nicer (IMO).


The emulation proposed by jcastromail (below) is useful to solution a frequent problem: the server's and client's clock synchronization. By setting and deleting cookies via Javascript, the browser will use the client's clock as basis and the cookie will expire at desired time, no matter if it is seven years in the past or in the future.

do not spam m. kristall

The "bug" that prevents cookies from working properly with Windows XP and Internet Explorer 6 seems to be a more strict following of the cookie specifications. This part in particular:
Only hosts within the specified domain can set a cookie for a domain and domains must have at least two (2) or three (3) periods in them to prevent domains of the form: ".com", ".edu", and "". Any domain that fails [sic] within one of the seven special top level domains listed below only require two periods. Any other domain requires at least three. The seven special top level domains are: "COM", "EDU", "NET", "ORG", "GOV", "MIL", and "INT".
localhost should not work. Neither should localhost.localdomain. Though, it may be a bug that does not work, even despit the fact that is not a domain name.

09-feb-2007 01:13

something that wasn't made clear to me here and totally confused me for a while was that domain names must contain at least two dots (.), hence 'localhost' is invalid and the browser will refuse to set the cookie! instead for localhost you should use false.
to make your code work on both localhost and a proper domain, you can do this:
$domain = ($_SERVER['HTTP_HOST'] != 'localhost') ? $_SERVER['HTTP_HOST'] : false;
setcookie('cookiename', 'data', time()+60*60*24*365, '/', $domain, false);


Some versions of IE6, when you drop a non-persistent cookie, handle this incorrectly and instead drop a persistent cookie.
This is not PHP's fault: ) Rather it is a bug in these versions of IE6. I am posting this because people may be programming session IDs and tokens in what they believe to be non-persistent cookies. Specifically, IE6 build 6.0.2800.1106CO (SP1) does this.
IE6, in this case, will make your application insecure. You should notify IE6 users to manually delete their cookies after their session is completed. Will they do this, no. At least you did your part to notify them of the risk.
We found this to be true after an ethical hack which condemned the use of SessionID's in query string, even if the entire session begins, happens, and ends, in https. M$ Proxy Server (here we go with M$ crappy security) writes these session ID's to it's proxy logs. Since Proxy Server is easily comprimised with numerous well known hacks, your sessions can be. I have seen the logs and seen the sessionid's in https.
Unfortunately, these EH fellows are forgetting that most user's systems are less secure and less patched than most ISP's and corporate proxy servers. We are simply moving the vulnerability to the client by using cookies. If it were up to me, I would leave sessionid in query string.
I was proven right in the case with this browser.


Setting the expiration time using strtotime is a very convenient way to calculate cookie expiration.
setcookie ("TestCookie", "", strtotime("+1 week"));


setcookie + header Location + IIS 5 = Trouble
It took me a long time to figure out what was causing a missing cookie in one system while it worked perfectly in another...
See this one:
In short, this WILL NEVER WORK IN IIS 5:
header("Pragma: no-cache");
setcookie('BB','2',time() + 24 * 3600,'/');
You will ONLY get the Location Header, everything else will be "cut out" by IIS 5 CGI implementation.
1- Migrate to Apache/IIS6/Whatever
2- Use a Non Parsed Header Script (nph-*.php)
3- Try with header('Refresh: 0; $URL');
I hope this helps somebody not to spend hours knocking his/her head.


Regarding to "rajneshavon@gmail dot com" note:
There only one type of cookies, simply -- cookie is the data storage itself, without any types/categories.
Cookies have some parameters (expiration date, domain, path) which should be considered.
Sessions mechanism works kinda simply:
when You call "session_start();" -- server attaches header
Set_cookie: session_name()=session_id(); path=/
usually. This cookie can be used in the next browser request to distinct one user from another (suggesting that session_id() has unique value for every user -- is't usually 32 md-hash).
To be completely simple -- server just tries to open file at session storage with the name like sess_<session_id()>. If that file exists -- server reads the data and performs unserialize() and that gives You opportunity to get the stored data as $_SESSION array -- this array is being simply serialized/unserialized when operating with it in scripts.
If browser disables cookies -- You can set
session.use_trans_sid = 1
That will tell Your PHP parser to append ?session_name()=session_id() to every link being outputted at the page and also add hidden field with that name=value at forms if any.
Upon clicking such link with session in the query string, You will receive session_id but from $_GET array. That will tell server where to get session data, just without cookies.
The only minus is that Your links will become "variable" with that boring session hash.


Problem with setcookie() and UTF-8 recognizing by browser:
- If you want to use UTF-8 characters in your php file, some editors insert special bytes in the very beginning of the file. This prevents setcookie() from working, because these special bytes are sent to the browser BEFORE the header, and you get "Header already sent" error;
- If you delete these bytes (with a hex editor), setcookie() will work fine, but the browser will STOP recognizing UTF-8 encoding automatically! The user will need to set the encoding to UTF-8 manually to see your page correctly.
Here's how to get out of this:
Instead of:
setcookie("aaa", "bbb");
<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
Write this:
header("Content-Type: text/html; charset=utf-8");
setcookie("aaa", "bbb");
(make sure you have no special bytes before "<?")


P3P is a good idea. But.
IE 6 features an inadequate definition of third party cookies.
If your site is hosted on server A and your PHP stuff is coming in a framesetting from server B your setcookie-attempts will be blocked when default privacy settings are deployed. Your secondparty-cookie will be regarded as a thirdparty-cookie.
So what you do is not read the P3P-Internet Explorer 6-manual at MS but send a header like
before doing the setcookie-thing. This will enable your cookie to survive any IE-6 privacy settings.
You won't do this if you're working for some doubleclick institution, because if you do, you... you...well... you are not a very nice person!


Observed what I think is a bug: session cookies were expiring even though the session was still active.  (To test, set a cookie expiry of 5 seconds and keep hitting the page every second.  The session will expire and create a new SESSID after 5 seconds despite the fact that you hit the page only a second ago.)
Calling this function before starting the session fixed it.  It copies the cookie contents back to itself while forcing an update to the expiry time in the cookie.
function FreshenSessionCookie($lifetimeSeconds, $cookieName = 'PHPSESSID')
if (isset($_COOKIE[$cookieName]))
$data = $_COOKIE[$cookieName];
$timeout = time()+$lifetimeSeconds;
setcookie($cookieName, $data,  $timeout);

paul nospam

Note when setting "array cookies" that a separate cookie is set for each element of the array.
On high traffic sites, this can substantially increase the size of subsequent HTTP requests from clients (including requests for static content on the same domain).
More importantly though, the cookie specification says that browsers need only accept 20 cookies per domain.  This limit is increased to 50 by Firefox, and to 30 by Opera, but IE6 and IE7 enforce the limit of 20 cookie per domain.  Any cookies beyond this limit will either knock out an older cookie or be ignored/rejected by the browser.

sijmen ruwhof

My experience with setting cookies, a must read for every developer that doenst want to spend his time debugging cookies..
- if your're using windows XP with Internet Explorer, having a local webserver running and you're setting a cookie in you script, don't set the domain name! Use false instead. Windows XP in combination with Internet Explorer (with Firefox your cookie will be set successful) will not set a cookie when the domain name points to localhost. Examples:
setcookie('test', 'value', false, '/', 'localhost', 0); // wil not work in WinXP with IE, but will work in WinXP with Firefox
Someone noted already this bug in WinXP + IE, but said that the domainname should have a dot in it. That is totally untrue (i've tried a domainname with a dot in it that points to I think that the problem persist that the domainname may not be pointing to so the solution would be:
setcookie('test', 'value', false, '/', false, 0); // works
- be sure that the 4th argument (path) always ends with a slash '/'
- be sure that the expire time is false or a timestamp, 0 is not the value for a session cookie (false is the value).
-- Sijmen Ruwhof

manfred dot p dot wilke

mrvanes at gmail dot com:
The note did you wrote about the use of underscore (_) on 06-Jun-2006 11:22 basicaly saved my life.
I was encountering some troubles with Windows_Xp and IE 6 and 7, because IE does not recognize a value like "aaabbb_cccddd" (i think).
After change the value, deleting the underscore, the cookie works fine.
This is the complete line that worked:
setcookie("value", date("Y-m-d"), time()+3600, "/", false, 0);
Tks very much.
(PS.: Sorry for my bad english.)


It would seem that set_cookie will still set a cookie via HTTP, even if the secure parameter is true.
So even if your site uses HTTPS/SSL to communicate sensitive data, if the initial cookie was set by set_cookie via HTTP, an attacker listening on the wire could easily spoof a visitor's cookie and gain access to their session.


Internet Exploer 6 now requires sites that set cookies to have P3P policies.
From the Microsoft page:
"Internet Explorer 6 implements advanced cookie filtering that is based on the Platform for Privacy Preferences (P3P) specification. By default, Internet Explorer 6 blocks third-party cookies that do not have a compact policy (a condensed computer-readable privacy statement) or third-party cookies that have a compact policy which specifies that personally identifiable information is used without your implicit consent. First-party cookies that have a compact policy which specifies that personally identifiable information is used without implicit consent are downgraded (deleted when you close Internet Explorer). First-party cookies that do not have a compact policy are leashed (restricted so that they can only be read in the first-party context)."
For more about P3P:


In response to Snupha's note, one way to get the desired result on the server side, and therefore work around the user's local date and time, is to store your session information in a sessions table in your database.
For example:
1. Create a sessions table in your favorite db with a 32 char sessionid column, modifed and created datetime columns, and any other columns that you want to associate with the session such as a userid.
2. In your PHP script, create a variable containing the 32 char unique ID for your session, and store the result in a sessionid cookie that expires at the end of the browser session, and also insert it in a new row in your sessions table.
3. When your scripts check for the existance of the sessionid cookie, if a value is returned, they would check against the sessions table in your database to find the matching ID, updating the modified datetime for that record to keep the session active.  If the server time elapsed since the record was modified exceeds your timeout constraints, you can force the user to reauthenticate and issue them a new session ID if desired.
4. If you require a login to some part of your site, you can also update the session record with the userid to link the session to the user.
I decided to use this method on a recent project where both PHP and ColdFusion needed to access the same datasource for different applications.  The result is a language-independent session in which multiple web servers on the same domain (but potentially different sub-domains) can know when the user has logged in and allow them to continue the session on the other server.  Otherwise, using the sessions built in to PHP and ColdFusion would require separate logins when switching between applications programmed in the two languages.
Going a little further with this, you can then report stats on the sessions table for the average session length by comparing the created and modified columns, or get a session history for a particular user.
I hope this helps a few people.  Have fun!


If you're looking to set multiple values in your cookie (rather than setting multiple cookies) you might find these useful.
function build_cookie($var_array) {
 if (is_array($var_array)) {
   foreach ($var_array as $index => $data) {
     $out.= ($data!="") ? $index."=".$data."|" : "";
 return rtrim($out,"|");
function break_cookie ($cookie_string) {
 foreach ($array as $i=>$stuff) {
 return $array;
Hopefully someone finds these useful.

carl v

If you want to delete all the cookies set by your domain, you may run the following:
$cookiesSet = array_keys($_COOKIE);
for ($x=0;$x<count($cookiesSet);$x++) setcookie($cookiesSet[$x],"",time()-1);
Very useful when doing logout scripts and the cookie name may have changed (long story).


If you want to delete all cookies on your domain, you may want to use the value of:
<?php $_SERVER['HTTP_COOKIE'] ?>
rather than:
<?php $_COOKIE ?>
to dertermine the cookie names.
If cookie names are in Array notation, eg: user[username]
Then PHP will automatically create a corresponding array in $_COOKIE. Instead use $_SERVER['HTTP_COOKIE'] as it mirrors the actual HTTP Request header.
// unset cookies
if (isset($_SERVER['HTTP_COOKIE'])) {
$cookies = explode(';', $_SERVER['HTTP_COOKIE']);
foreach($cookies as $cookie) {
$parts = explode('=', $cookie);
$name = trim($parts[0]);
setcookie($name, '', time()-1000);
setcookie($name, '', time()-1000, '/');


If you want to delete a session cookie, you can do it with this code:
 // many code
 $sessionName = session_name();
 $sessionCookie = session_get_cookie_params();
 setcookie($sessionName, false, $sessionCookie['lifetime'], $sessionCookie['path'], $sessionCookie['domain'], $sessionCookie['secure']);
This works also well if the session cookie params or the session name were changed.

21-mar-2007 10:40

if you only want to do something once per unique visitor, you can test if a cookie is set, and if not, set the cookie and perform the action. This being the poorman's version, it has a problem, where if a user is blocking cookies they will appear as a first time visitor each time. What you can do to avoid this is to set a test cookie first and check that it exists. If it exists, then check to see if your second cookie has been set. If the first one is set, but the second isn't, then you know this is a first time visitor.


If you intend to use persistent cookies (vice session cookies that are deleted when the browser is closed) be aware:
1) Firefox appears to require that you include all paramaters, or it will ignore the expiration and treat the cookie as a session cookie
2) My version of firefox ( defaults to 'keep cookies until i close firefox' , which essentially makes every cookie a session cookie. This of course sucks for devs, but i suppose is supposed to be a security feature for the end user. If the user wants to configure firefox to respect the expiration date and retain cookies beyond the session, the user must change it to 'keep cookies until they expire'.


If you experience problems on Microsoft Information Server (IIS) when setting a cookie via PHP and when PHP is running as a CGI binary it is not setting the cookie. After spending many hours on what the problem is here is what happens:
When you invoke setcookie and redirect to another page you will not have your cookie set, because it seems that IIS doesn't set the cookie unless you actually send some contents to the browser. Therefore if you want to set a cookie and then redirect to another page you will have to do the redirection via JavaScript and/or HTML if you want your script to work on IIS.


If you ever have to modify, add, or delete cookies (that you added with php) using Javascript, try using this piece of code i found here:
Its rather simple and very useful.

05-jan-2007 09:33

If you ever find yourself in a situation where you need to overwrite a non-PHP application's session cookie, you can do that with the following line:
header("Set-Cookie: SIDNAME=$overwrite; path=/; secure");
I couldn't get setcookie() to do this for all major web browsers, but manually sending the header did the trick. Note: Remove secure if you aren't mandating SSL connections.


If you cannot get a cookie to stick when using an expiration date - check that local computer's time and time zone!
I spent hours trouble-shooting my code when I realized that the computer time on a client's computer was 7 years into the future which automatically expired the cookie before it could be used! Go figure, I'm going to look now for a work around that I can do on the server-side.


if you are having problems seeing cookies sometimes or deleting cookies sometimes, despite following the advice below, make sure you are setting the cookie with the domain argument. Set it with the dot before the domain as the examples show: "".  I wasn't specifying the domain, and finally realized I was setting the cookie when the browser url had the and later trying to delete it when the url didn't have the www. ie. This also caused the page to be unable to find the cookie when the www. wasn't in the domain.  (When you add the domain argument to the setcookie code that creates the cookie, make sure you also add it to the code that deletes the cookie.)


If you are having issues with IE7 and setcookie(), be sure to verify that the cookie is set via http for http sites, and https for https site.
Also, if the time is incorrect on your server, IE7 will also disallow those cookies from being set.


I've just been fighting with an issue with setting cookie paths that contain the underscore [_] and dash [-] characters on IIS6.
This issue is a known "feature" that was added to IIS6 to help prevent cross-site scripting vunerabilities (See:
Basically, when IIS6 is asked to go to a URI that doesn't contain a trailing slash, it will "courtesy redirect" to the URI with a trailing slash, but in doing so will encode underscores and dashes into their ASCII codes (%5F and %2D respectively).
Now, while this normally wouldn't be an issue, when a cookie is set by PHP using a defined path the unencoded version is used for the path. PHP does not know that the path is encoded, as IIS6 decodes it before it's given to the script (I assume).
IIS redirects =
PHP sees:
PHP_SELF: /some_directory/script.php
So, setting the path: path="/some_directory/"
but browser sees: path="/some%5Fdirectory/" which are not equal, so therefore the paths aren't the same and the cookie isn't retrieved.
Normally this wouldn't be an issue: simply don't specify a path & the browser handles it for you; or don't use underscores or dashes in the URI; or specify the URI with the trailing slash. But when used to lock down sessions to a specific directory (which might have those characters) this becomes a big problem (default session path is /).
The solution is to also set a cookie using the session information (name, value & other parameters) for the encoded version of the path, then the same session will be used for either version of the path.


I've found that when using header("Location: <destination page>") type 302 redirect, that cookie information in the header will be silently dropped if the <destination page> path includes any non-alphanumeric (except the allowed "-" and ".") characters. That had been pointed out above, and is documented with Microsoft as by design...;EN-US;316112
...but what is not made clear, is that the rule seems to include file and directory names too!
Also, I've had unpredictable results when the <destination page> filename begins with ANY character other than strictly alphanumeric characters, even the otherwise allowed "-" and "." characters; and/or if the <destination page> is in a different directory.

robert chapin

I would enjoy any new comments or documentation to address these compliance problems:
RFC 2109 defines the cookie VALUE as a quoted-string.
RFC 2965 obsoletes the Set-Cookie header.
Robert Chapin
Chapin Information Services


I just discovered that IE6 (6.0.2800.1106 at least) seems to have difficulties accepting cookies when the subdomain part of the URL is longer than 10 characters. This applies to automatic sessionid's as well (that's how I found out). So: Setting a cookie on a site named will work, doing the same on will fail (or at least, failed for me).
Tested it by snooping the outgoing traffic from the client: the GET request lacks the COOKIE: header in the last case. A closer inspection revealed that the cookie was never accepted in the first place: when the subdomain is longer than 10 characters, no cookiefile is created in the temporary internet files directory. Temporary cookies (like the PHP sessionid) have the same problem.
Firefox and Konqueror have no problem with this.
[Editor's note:  Your comment is wrong. It is not the length of the hostname but the use of the underscore (_) char within it which makes IE (silently) freak out. The use of underscores in hostnames is an RFC violation and may very well not be supported at all by a browser, proxy or any http client.]


I found out recently that assigning FALSE to a cookie will destroy it.
I thought it might interest some of you.


httponly attribute, added in 5.2 works only in IE

brian dot powell

Here is problem I ran into during a recent bout with IE7 and cookies.  IE will not delete a cookie value if the time is set to the past.  It will hold the value no matter how far in the past you set the "expire" value.  IE7 is the only browser I have had problems with - so here is the solution I came up with.
//check to see how to set the cookie
$Browsertype = $_SERVER['HTTP_USER_AGENT'];
$Parts = explode(" ",$Browsertype);
$MSIE = array_search("MSIE",$Parts);

setcookie("name", "", time()+20000);
setcookie("name", "", time()-20000, "/", "" );


great firefox extension to enable httpOnly


Firefox is following the real spec and does not decode '+' to fact it further encodes them to '%2B' to store the cookie.  If you read a cookie using javascript and unescape it, all your spaces will be turned to '+'.
To fix this problem, use setrawcookie and rawurlencode:
setrawcookie('cookie_name', rawurlencode($value), time()+60*60*24*365);
The only change is that spaces will be encoded to '%20' instead of '+' and will now decode properly.


Browsers like Opera and Mozilla are told to need the full page loaded to accept cookies (that's what bmatzelle at NOSPAM dot yahoo dot com said).
You can use cookies at reloading pages, however after sending out all headers call exit(); to stop the rest of the script, so the actual page content is not sent out.


Be careful of using the same cookie name in subdirectories. Setting a simple cookie
<?setcookie("region", $_GET['set_region']);?>
both in the root / and for instance in this case /admin/ will create 2 cookies with different paths. In reading the cookies back only the first one is read regardless of path.


As per Sijmen Ruwhof's comment:
It's not that the domain requires a dot as much as it requires a fully qualified domain name.  Setting it to false is a valid workaround, as it will default to whatever domain the page is being accessed at.  If a fully qualified domain name is unavailable, consider editing your hosts file on both the client and server to have a common domain name to use for the cookie.


An undocumented "feature" of setcookie:
setcookie( 'delete_me', '', time() - 1000, '/', '', '' );
The setcookie function will silently change the '' cookie value to 'deleted' and override your expiry time to time() - 1 year.
I only noticed this when the computer I was testing on had it's date set incorrectly to the > 1 year in the past.  The computer connects to the server which has it's date correctly set and thus the cookie on the client doesn't delete but stores the value "deleted" instead.


About the IIS bug (cookies removed in CGI if 302 Redirect)
reported by shrockc (2003-12-23), here's the correct URL.
BUG: Set-Cookie Is Ignored in CGI When Combined With Location

jonathan dot bergeron

About the delete part, I found that Firefox only remove the cookie when you submit the same values for all parameters, except the date, which sould be in the past. Submiting blank values didn't work for me.
Example :
- set -
setcookie( "name", "value", "future_timestamp", "path", "domain" );
- delete -
setcookie( "name", "value", "past_timestamp", "path", "domain" );


A few comments have suggested using serialize() to set object or array data into a cookie.  There are a couple of reasons to be carefull with that technique:
Security: If the cookie is human readable, then it is also fairly easy for end users to play around with it.  Wrapping your cookie setting and getting in an encryption routine will prevent tampering, and make sure that your cookies don't make any sense to any client-side exploits or other sites they get sent to thanks to browser bugs.
Bulk: If you serialize even a fairly simple class, then you get a lot of data.  Large cookies will make browser requests fat and slow, and some browsers have a limit on cookie size, so think about what data you really need to persist, and create __sleep() and __wakeup() methods to package the data into the shortest possible form.  You can get better and faster results when you write your own __sleep() and __wakup() to implode() or pack() your data, than by using zlib compress() on the serialized object.


// to create a session cookie use zero (0) instead of time
setcookie("name", "value", 0);
// to destroy the session-cookie use
setcookie("name", "", -1);

marcin wiazowski

'session.cookie_domain' should be set to empty string for all local domain names, not only for 'localhost' (but should not be empty for local IP addresses):
ini_set('session.cookie_domain', (strpos($_SERVER['HTTP_HOST'],'.') !== false) ? $_SERVER['HTTP_HOST'] : '');

Change Language

Follow Navioo On Twitter
eXTReMe Tracker