Configuring repoze.who

Configuration Points


repoze.who “classifies” the request on middleware ingress. Request classification happens before identification and authentication. A request from a browser might be classified a different way than a request from an XML-RPC client. repoze.who uses request classifiers to decide which other components to consult during subsequent identification, authentication, and challenge steps. Plugins are free to advertise themselves as willing to participate in identification and authorization for a request based on this classification. The request classification system is pluggable. repoze.who provides a default classifier that you may use.

You may extend the classification system by making repoze.who aware of a different request classifier implementation.

Challenge Deciders

repoze.who uses a “challenge decider” to decide whether the response returned from a downstream application requires a challenge plugin to fire. When using the default challenge decider, only the status is used (if it starts with 401, a challenge is required).

repoze.who also provides an alternate challenge decider, repoze.who.classifiers.passthrough_challenge_decider, which avoids challenging 401 responses which have been “pre-challenged” by the application.

You may supply a different challenge decider as necessary.


repoze.who has core functionality designed around the concept of plugins. Plugins are instances that are willing to perform one or more identification- and/or authentication-related duties. Each plugin can be configured arbitrarily.

repoze.who consults the set of configured plugins when it intercepts a WSGI request, and gives some subset of them a chance to influence what repoze.who does for the current request.


As of repoze.who 1.0.7, the repoze.who.plugins package is a namespace package, intended to make it possible for people to ship eggs which are who plugins as, e.g. repoze.who.plugins.mycoolplugin.

Configuring repoze.who via Python Code

class repoze.who.middleware.PluggableAuthenticationMiddleware(app, identifiers, challengers, authenticators, mdproviders, classifier, challenge_decider[, log_stream=None[, log_level=logging.INFO[, remote_user_key='REMOTE_USER']]])

The primary method of configuring the repoze.who middleware is to use straight Python code, meant to be consumed by frameworks which construct and compose middleware pipelines without using a configuration file.

In the middleware constructor: app is the “next” application in the WSGI pipeline. identifiers is a sequence of IIdentifier plugins, challengers is a sequence of IChallenger plugins, mdproviders is a sequence of IMetadataProvider plugins. Any of these can be specified as the empty sequence. classifier is a request classifier callable, challenge_decider is a challenge decision callable. log_stream is a stream object (an object with a write method) or a logging.Logger object, log_level is a numeric value that maps to the logging module’s notion of log levels, remote_user_key is the key in which the REMOTE_USER (userid) value should be placed in the WSGI environment for consumption by downstream applications.

An example configuration which uses the default plugins follows:

from repoze.who.middleware import PluggableAuthenticationMiddleware
from repoze.who.interfaces import IIdentifier
from repoze.who.interfaces import IChallenger
from repoze.who.plugins.basicauth import BasicAuthPlugin
from repoze.who.plugins.auth_tkt import AuthTktCookiePlugin
from repoze.who.plugins.redirector import RedirectorPlugin
from repoze.who.plugins.htpasswd import HTPasswdPlugin

io = StringIO()
salt = 'aa'
for name, password in [ ('admin', 'admin'), ('chris', 'chris') ]:
    io.write('%s:%s\n' % (name, password))
def cleartext_check(password, hashed):
    return password == hashed
htpasswd = HTPasswdPlugin(io, cleartext_check)
basicauth = BasicAuthPlugin('repoze.who')
auth_tkt = AuthTktCookiePlugin('secret', 'auth_tkt', digest_algo="sha512")
redirector = RedirectorPlugin('/login.html')
redirector.classifications = {IChallenger:['browser'],} # only for browser
identifiers = [('auth_tkt', auth_tkt),
               ('basicauth', basicauth)]
authenticators = [('auth_tkt', auth_tkt),
                  ('htpasswd', htpasswd)]
challengers = [('redirector', redirector),
               ('basicauth', basicauth)]
mdproviders = []

from repoze.who.classifiers import default_request_classifier
from repoze.who.classifiers import default_challenge_decider
log_stream = None
import os
if os.environ.get('WHO_LOG'):
    log_stream = sys.stdout

middleware = PluggableAuthenticationMiddleware(
    log_stream = log_stream,
    log_level = logging.DEBUG

The above example configures the repoze.who middleware with:

  • Two IIdentifier plugins (auth_tkt cookie, and a basic auth plugin). In this setup, when “identification” needs to be performed, the auth_tkt plugin will be checked first, then the basic auth plugin. The application is responsible for handling login via a form: this view would use the API (via :method:`remember`) to generate apprpriate response headers.
  • Two IAuthenticator plugins: the auth_tkt plugin and an htpasswd plugin. The auth_tkt plugin performs both IIdentifier and IAuthenticator functions. The htpasswd plugin is configured with two valid username / password combinations: chris/chris, and admin/admin. When an username and password is found via any identifier, it will be checked against this authenticator.
  • Two IChallenger plugins: the redirector plugin, then the basic auth plugin. The redirector auth will fire if the request is a browser request, otherwise the basic auth plugin will fire.

The rest of the middleware configuration is for values like logging and the classifier and decider implementations. These use the “stock” implementations.


The app referred to in the example is the “downstream” WSGI application that who is wrapping.

Configuring repoze.who via Config File

repoze.who may be configured using a ConfigParser-style .INI file. The configuration file has five main types of sections: plugin sections, a general section, an identifiers section, an authenticators section, and a challengers section. Each “plugin” section defines a configuration for a particular plugin. The identifiers, authenticators, and challengers sections refer to these plugins to form a site configuration. The general section is general middleware configuration.

To configure repoze.who in Python, using an .INI file, call the make_middleware_with_config entry point, passing the right-hand application, the global configuration dictionary, and the path to the config file. The global configuration dictionary is a dictonary passed by PasteDeploy. The only key ‘make_middleware_with_config’ needs is ‘here’ pointing to the config file directory. For debugging people might find it useful to enable logging by adding the log_file argument, e.g. log_file=”repoze_who.log”

from repoze.who.config import make_middleware_with_config
global_conf = {"here": "."}  # if this is not defined elsewhere
who = make_middleware_with_config(app, global_conf, 'who.ini')

repoze.who’s configuration file can be pointed to within a PasteDeploy configuration file

use = egg:repoze.who#config
config_file = %(here)s/who.ini
log_file = stdout
log_level = debug

Below is an example of a configuration file (what config_file might point at above ) that might be used to configure the repoze.who middleware. A set of plugins are defined, and they are referred to by following non-plugin sections.

In the below configuration, five plugins are defined. The form, and basicauth plugins are nominated to act as challenger plugins. The form, cookie, and basicauth plugins are nominated to act as identification plugins. The htpasswd and sqlusers plugins are nominated to act as authenticator plugins.

# identificaion and challenge
use = repoze.who.plugins.redirector:make_plugin
login_url = /login.html

# identification and authentication
use = repoze.who.plugins.auth_tkt:make_plugin
secret = s33kr1t
cookie_name = oatmeal
secure = False
include_ip = False
digest_algo = sha512

# identification and challenge
use = repoze.who.plugins.basicauth:make_plugin
realm = 'sample'

# authentication
use = repoze.who.plugins.htpasswd:make_plugin
filename = %(here)s/passwd
check_fn = repoze.who.plugins.htpasswd:crypt_check

# authentication
use = repoze.who.plugins.sql:make_authenticator_plugin
# Note the double %%:  we have to escape it from the config parser in
# order to preserve it as a template for the psycopg2, whose 'paramstyle'
# is 'pyformat'.
query = SELECT userid, password FROM users where login = %%(login)s
conn_factory = repoze.who.plugins.sql:make_psycopg_conn_factory
compare_fn = repoze.who.plugins.sql:default_password_compare

name = properties
use = repoze.who.plugins.sql:make_metadata_plugin
# Note the double %%:  we have to escape it from the config parser in
# order to preserve it as a template for the psycopg2, whose 'paramstyle'
# is 'pyformat'.
query = SELECT firstname, lastname FROM users where userid = %%(__userid)s
filter = my.package:filter_propmd
conn_factory = repoze.who.plugins.sql:make_psycopg_conn_factory

request_classifier = repoze.who.classifiers:default_request_classifier
challenge_decider = repoze.who.classifiers:default_challenge_decider
remote_user_key = REMOTE_USER

# plugin_name;classifier_name:.. or just plugin_name (good for any)
plugins =

# plugin_name;classifier_name.. or just plugin_name (good for any)
plugins =

# plugin_name;classifier_name:.. or just plugin_name (good for any)
plugins =

plugins =

The basicauth section configures a plugin that does identification and challenge for basic auth credentials. The redirector section configures a plugin that does challenges. The auth_tkt section configures a plugin that does identification for cookie auth credentials, as well as authenticating them. The htpasswd plugin obtains its user info from a file. The sqlusers plugin obtains its user info from a Postgres database.

The identifiers section provides an ordered list of plugins that are willing to provide identification capability. These will be consulted in the defined order. The tokens on each line of the plugins= key are in the form “plugin_name;requestclassifier_name:…” (or just “plugin_name” if the plugin can be consulted regardless of the classification of the request). The configuration above indicates that the system will look for credentials using the auth_tkt cookie identifier (unconditionally), then the basic auth plugin (unconditionally).

The authenticators section provides an ordered list of plugins that provide authenticator capability. These will be consulted in the defined order, so the system will look for users in the file, then in the sql database when attempting to validate credentials. No classification prefixes are given to restrict which of the two plugins are used, so both plugins are consulted regardless of the classification of the request. Each authenticator is called with each set of identities found by the identifier plugins. The first identity that can be authenticated is used to set REMOTE_USER.

The mdproviders section provides an ordered list of plugins that provide metadata provider capability. These will be consulted in the defined order. Each will have a chance (on ingress) to provide add metadata to the authenticated identity. Our example mdproviders section shows one plugin configured: “sqlproperties”. The sqlproperties plugin will add information related to user properties (e.g. first name and last name) to the identity dictionary.

The challengers section provides an ordered list of plugins that provide challenger capability. These will be consulted in the defined order, so the system will consult the cookie auth plugin first, then the basic auth plugin. Each will have a chance to initiate a challenge. The above configuration indicates that the redirector challenger will fire if it’s a browser request, and the basic auth challenger will fire if it’s not (fallback).