Zend_Auth_Adapter_Ldap supports web application authentication with LDAP services. Its features include username and domain name canonicalization, multi-domain authentication, and failover capabilities. It has been tested to work with » Microsoft Active Directory and » OpenLDAP, but it should also work with other LDAP service providers.

This documentation includes a guide on using Zend_Auth_Adapter_Ldap, an exploration of its API, an outline of the various available options, diagnostic information for troubleshooting authentication problems, and example options for both Active Directory and OpenLDAP servers.

To incorporate Zend_Auth_Adapter_Ldap authentication into your application quickly, even if you're not using Zend_Controller, the meat of your code should look something like the following:

Of course, the logging code is optional, but it is highly recommended that you use a logger. Zend_Auth_Adapter_Ldap will record just about every bit of information anyone could want in $messages (more below), which is a nice feature in itself for something that has a history of being notoriously difficult to debug.

The Zend_Config_Ini code is used above to load the adapter options. It is also optional. A regular array would work equally well. The following is an example application/config/config.ini file that has options for two separate servers. With multiple sets of server options the adapter will try each, in order, until the credentials are successfully authenticated. The names of the servers (e.g., 'server1' and 'server2') are largely arbitrary. For details regarding the options array, see the Server Options section below. Note that Zend_Config_Ini requires that any values with "equals" characters (=) will need to be quoted (like the DNs shown below).

The above configuration will instruct Zend_Auth_Adapter_Ldap to attempt to authenticate users with the OpenLDAP server s0.foo.net first. If the authentication fails for any reason, the AD server dc1.w.net will be tried.

With servers in different domains, this configuration illustrates multi-domain authentication. You can also have multiple servers in the same domain to provide redundancy.

Note that in this case, even though OpenLDAP has no need for the short NetBIOS style domain name used by Windows, we provide it here for name canonicalization purposes (described in the Username Canonicalization section below).

The Zend_Auth_Adapter_Ldap constructor accepts three parameters.

The $options parameter is required and must be an array containing one or more sets of options. Note that it is an array of arrays of Zend_Ldap options. Even if you will be using only one LDAP server, the options must still be within another array.

Below is » print_r() output of an example options parameter containing two sets of server options for LDAP servers s0.foo.net and dc1.w.net (the same options as the above INI representation):

The information provided in each set of options above is different mainly because AD does not require a username be in DN form when binding (see the bindRequiresDn option in the Server Options section below), which means we can omit a number of options associated with retrieving the DN for a username being authenticated.

Note: What is a Distinguished Name?
A DN or "distinguished name" is a string that represents the path to an object within the LDAP directory. Each comma-separated component is an attribute and value representing a node. The components are evaluated in reverse. For example, the user account CN=Bob Carter,CN=Users,DC=w,DC=net is located directly within the CN=Users,DC=w,DC=net container. This structure is best explored with an LDAP browser like the ADSI Edit MMC snap-in for Active Directory or phpLDAPadmin.

The names of servers (e.g. 'server1' and 'server2' shown above) are largely arbitrary, but for the sake of using Zend_Config, the identifiers should be present (as opposed to being numeric indexes) and should not contain any special characters used by the associated file formats (e.g. the '.' INI property separator, '&' for XML entity references, etc).

With multiple sets of server options, the adapter can authenticate users in multiple domains and provide failover so that if one server is not available, another will be queried.

Note: The Gory Details: What Happens in the Authenticate Method?
When the authenticate() method is called, the adapter iterates over each set of server options, sets them on the internal Zend_Ldap instance, and calls the Zend_Ldap::bind() method with the username and password being authenticated. The Zend_Ldap class checks to see if the username is qualified with a domain (e.g., has a domain component like [email protected] or FOO\alice). If a domain is present, but does not match either of the server's domain names (foo.net or FOO), a special exception is thrown and caught by Zend_Auth_Adapter_Ldap that causes that server to be ignored and the next set of server options is selected. If a domain does match, or if the user did not supply a qualified username, Zend_Ldap proceeds to try to bind with the supplied credentials. if the bind is not successful, Zend_Ldap throws a Zend_Ldap_Exception which is caught by Zend_Auth_Adapter_Ldap and the next set of server options is tried. If the bind is successful, the iteration stops, and the adapter's authenticate() method returns a successful result. If all server options have been tried without success, the authentication fails, and authenticate() returns a failure result with error messages from the last iteration.

The username and password parameters of the Zend_Auth_Adapter_Ldap constructor represent the credentials being authenticated (i.e., the credentials supplied by the user through your HTML login form). Alternatively, they may also be set with the setUsername() and setPassword() methods.

Each set of server options in the context of Zend_Auth_Adapter_Ldap consists of the following options, which are passed, largely unmodified, to Zend_Ldap::setOptions():

Server Options

Name	Description
host	The hostname of LDAP server that these options represent. This option is required.
port	The port on which the LDAP server is listening. If useSsl is TRUE, the default port value is 636. If useSsl is FALSE, the default port value is 389.
useStartTls	Whether or not the LDAP client should use TLS (aka SSLv2) encrypted transport. A value of TRUE is strongly favored in production environments to prevent passwords from be transmitted in clear text. The default value is FALSE, as servers frequently require that a certificate be installed separately after installation. The useSsl and useStartTls options are mutually exclusive. The useStartTls option should be favored over useSsl but not all servers support this newer mechanism.
useSsl	Whether or not the LDAP client should use SSL encrypted transport. The useSsl and useStartTls options are mutually exclusive, but useStartTls should be favored if the server and LDAP client library support it. This value also changes the default port value (see port description above).
username	The DN of the account used to perform account DN lookups. LDAP servers that require the username to be in DN form when performing the "bind" require this option. Meaning, if bindRequiresDn is TRUE, this option is required. This account does not need to be a privileged account; an account with read-only access to objects under the baseDn is all that is necessary (and preferred based on the Principle of Least Privilege).
password	The password of the account used to perform account DN lookups. If this option is not supplied, the LDAP client will attempt an "anonymous bind" when performing account DN lookups.
bindRequiresDn	Some LDAP servers require that the username used to bind be in DN form like CN=Alice Baker,OU=Sales,DC=foo,DC=net (basically all servers except AD). If this option is TRUE, this instructs Zend_Ldap to automatically retrieve the DN corresponding to the username being authenticated, if it is not already in DN form, and then re-bind with the proper DN. The default value is FALSE. Currently only Microsoft Active Directory Server (ADS) is known not to require usernames to be in DN form when binding, and therefore this option may be FALSE with AD (and it should be, as retrieving the DN requires an extra round trip to the server). Otherwise, this option must be set to TRUE (e.g. for OpenLDAP). This option also controls the default acountFilterFormat used when searching for accounts. See the accountFilterFormat option.
baseDn	The DN under which all accounts being authenticated are located. This option is required. if you are uncertain about the correct baseDn value, it should be sufficient to derive it from the user's DNS domain using DC= components. For example, if the user's principal name is [email protected], a baseDn of DC=foo,DC=net should work. A more precise location (e.g., OU=Sales,DC=foo,DC=net) will be more efficient, however.
accountCanonicalForm	A value of 2, 3 or 4 indicating the form to which account names should be canonicalized after successful authentication. Values are as follows: 2 for traditional username style names (e.g., alice), 3 for backslash-style names (e.g., FOO\alice) or 4 for principal style usernames (e.g., [email protected]). The default value is 4 (e.g., [email protected]). For example, with a value of 3, the identity returned by Zend_Auth_Result::getIdentity() (and Zend_Auth::getIdentity(), if Zend_Auth was used) will always be FOO\alice, regardless of what form Alice supplied, whether it be alice, [email protected], FOO\alice, FoO\aLicE, foo.net\alice, etc. See the Account Name Canonicalization section in the Zend_Ldap documentation for details. Note that when using multiple sets of server options it is recommended, but not required, that the same accountCanonicalForm be used with all server options so that the resulting usernames are always canonicalized to the same form (e.g., if you canonicalize to EXAMPLE\username with an AD server but to [email protected] with an OpenLDAP server, that may be awkward for the application's high-level logic).
accountDomainName	The FQDN domain name for which the target LDAP server is an authority (e.g., example.com). This option is used to canonicalize names so that the username supplied by the user can be converted as necessary for binding. It is also used to determine if the server is an authority for the supplied username (e.g., if accountDomainName is foo.net and the user supplies [email protected], the server will not be queried, and a failure will result). This option is not required, but if it is not supplied, usernames in principal name form (e.g., [email protected]) are not supported. It is strongly recommended that you supply this option, as there are many use-cases that require generating the principal name form.
accountDomainNameShort	The 'short' domain for which the target LDAP server is an authority (e.g., FOO). Note that there is a 1:1 mapping between the accountDomainName and accountDomainNameShort. This option should be used to specify the NetBIOS domain name for Windows networks, but may also be used by non-AD servers (e.g., for consistency when multiple sets of server options with the backslash style accountCanonicalForm). This option is not required but if it is not supplied, usernames in backslash form (e.g., FOO\alice) are not supported.
accountFilterFormat	The LDAP search filter used to search for accounts. This string is a » printf()-style expression that must contain one '%s' to accomodate the username. The default value is '(&(objectClass=user)(sAMAccountName=%s))', unless bindRequiresDn is set to TRUE, in which case the default is '(&(objectClass=posixAccount)(uid=%s))'. For example, if for some reason you wanted to use bindRequiresDn = true with AD you would need to set accountFilterFormat = '(&(objectClass=user)(sAMAccountName=%s))'.
optReferrals	If set to TRUE, this option indicates to the LDAP client that referrals should be followed. The default value is FALSE.

Note: If you enable useStartTls = TRUE or useSsl = TRUE you may find that the LDAP client generates an error claiming that it cannot validate the server's certificate. Assuming the PHP LDAP extension is ultimately linked to the OpenLDAP client libraries, to resolve this issue you can set "TLS_REQCERT never" in the OpenLDAP client ldap.conf (and restart the web server) to indicate to the OpenLDAP client library that you trust the server. Alternatively, if you are concerned that the server could be spoofed, you can export the LDAP server's root certificate and put it on the web server so that the OpenLDAP client can validate the server's identity.

Zend_Auth_Adapter_Ldap collects debugging information within its authenticate() method. This information is stored in the Zend_Auth_Result object as messages. The array returned by Zend_Auth_Result::getMessages() is described as follows

Debugging Messages

Messages Array Index	Description
Index 0	A generic, user-friendly message that is suitable for displaying to users (e.g., "Invalid credentials"). If the authentication is successful, this string is empty.
Index 1	A more detailed error message that is not suitable to be displayed to users but should be logged for the benefit of server operators. If the authentication is successful, this string is empty.
Indexes 2 and higher	All log messages in order starting at index 2.

In practice, index 0 should be displayed to the user (e.g., using the FlashMessenger helper), index 1 should be logged and, if debugging information is being collected, indexes 2 and higher could be logged as well (although the final message always includes the string from index 1).