Provided by: python-fedora-doc_0.10.0-1_all 
NAME
python-fedora - python-fedora 0.10.0
The python-fedora module makes it easier for programmers to create both Fedora Services and clients that
talk to the services. It's licensed under the GNU Lesser General Public License version 2 or later.
FEDORA CLIENT
Authors
Toshio Kuratomi Luke Macken
Date 28 May 2008
For Version
0.3.x
The client module allows you to easily code an application that talks to a Fedora Service. It handles
the details of decoding the data sent from the Service into a python data structure and raises an
Exception if an error is encountered.
BaseClient
The BaseClient class is the basis of all your interactions with the server. It is flexible enough to be
used as is for talking with a service but is really meant to be subclassed and have methods written for
it that do the things you specifically need to interact with the Fedora Service you care about. Authors
of a Fedora Service are encouraged to provide their own subclasses of BaseClient that make it easier for
other people to use a particular Service out of the box.
Using Standalone
If you don't want to subclass, you can use BaseClient as a utility class to talk to any Fedora Service.
There's three steps to this. First you import the BaseClient and Exceptions from the fedora.client
module. Then you create a new BaseClient with the URL that points to the root of the Fedora Service
you're interacting with. Finally, you retrieve data from a method on the server. Here's some code that
illustrates the process:
from fedora.client import BaseClient, AppError, ServerError
client = BaseClient('https://admin.fedoraproject.org/pkgdb')
try:
collectionData = client.send_request('/collections', auth=False)
except ServerError as e:
print('%s' % e)
except AppError as e:
print('%s: %s' % (e.name, e.message))
for collection in collectionData['collections']:
print('%s %s' % (collection['name'], collection['version'])
BaseClient Constructor
In our example we only provide BaseClient() with the URL fragment it uses as the base of all requests.
There are several more optional parameters that can be helpful.
If you need to make an authenticated request you can specify the username and password to use when you
construct your BaseClient using the username and password keyword arguments. If you do not use these,
authenticated requests will try to connect via a cookie that was saved from previous runs of BaseClient.
If that fails as well, BaseClient will throw an Exception which you can catch in order to prompt for a
new username and password:
from fedora.client import BaseClient, AuthError
import getpass
MAX_RETRIES = 5
client = BaseClient('https://admin.fedoraproject.org/pkgdb',
username='foo', password='bar')
# Note this is simplistic. It only prompts once for another password.
# Your application may want to loop through this several times.
while (count < MAX_RETRIES):
try:
collectionData = client.send_request('/collections', auth=True)
except AuthError as e:
client.password = getpass.getpass('Retype password for %s: ' % username)
else:
# data retrieved or we had an error unrelated to username/password
break
count = count + 1
WARNING:
Note that although you can set the username and password as shown above you do have to be careful in
cases where your application is multithreaded or simply processes requests for more than one user with
the same BaseClient. In those cases, you can accidentally overwrite the username and password between
two requests. To avoid this, make sure you instantiate a separate BaseClient for every thread of
control or for every request you handle or use ProxyClient instead.
The useragent parameter is useful for identifying in log files that your script is calling the server
rather than another. The default value is Fedora BaseClient/VERSION where VERSION is the version of the
BaseClient module. If you want to override this just give another string to this:
client = BaseClient('https://admin.fedoraproject.org/pkgdb',
useragent='Package Database Client/1.0')
The debug parameter turns on a little extra output when running the program. Set it to true if you're
having trouble and want to figure out what is happening inside of the BaseClient code.
send_request()
send_request() is what does the heavy lifting of making a request of the server, receiving the reply, and
turning that into a python dictionary. The usage is pretty straightforward.
The first argument to send_request() is method. It contains the name of the method on the server. It
also has any of the positional parameters that the method expects (extra path information interpreted by
the server for those building non-TurboGears applications).
The auth keyword argument is a boolean. If True, the session cookie for the user is sent to the server.
If this fails, the username and password are sent. If that fails, an Exception is raised that you can
handle in your code.
req_params contains a dictionary of additional keyword arguments for the server method. These would be
the names and values returned via a form if it was a CGI. Note that parameters passed as extra path
information should be added to the method argument instead.
An example:
import BaseClient
client = BaseClient('https://admin.fedoraproject.org/pkgdb/')
client.send_request('/package/name/python-fedora', auth=False,
req_params={'collectionVersion': '9', 'collectionName': 'Fedora'})
In this particular example, knowing how the server works, /packages/name/ defines the method that the
server is going to invoke. python-fedora is a positional parameter for the name of the package we're
looking up. auth=False means that we'll try to look at this method without having to authenticate. The
req_params sends two additional keyword arguments: collectionName which specifies whether to filter on a
single distro or include Fedora, Fedora EPEL, Fedora OLPC, and Red Hat Linux in the output and
collectionVersion which specifies which version of the distribution to output for.
The URL constructed by BaseClient to the server could be expressed as[#]_:
https://admin.fedoraproject.org/pkgdb/package/name/python-fedora/?collectionName=Fedora&collectionVersion=9
In previous releases of python-fedora, there would be one further query parameter: tg_format=json. That
parameter instructed the server to return the information as JSON data instead of HTML. Although this is
usually still supported in the server, BaseClient has deprecated this method. Servers should be
configured to use an Accept header to get this information instead. See the JSON output section of the
Fedora Service documentation for more information about the server side.
Subclassing
Building a client using subclassing builds on the information you've already seen inside of BaseClient.
You might want to use this if you want to provide a module for third parties to access a particular
Fedora Service. A subclass can provide a set of standard methods for calling the server instead of
forcing the user to remember the URLs used to access the server directly.
Here's an example that turns the previous calls into the basis of a python API to the Fedora Package
Database:
import getpass
import sys
from fedora.client import BaseClient, AuthError
class MyClient(BaseClient):
def __init__(self, baseURL='https://admin.fedoraproject.org/pkgdb',
username=None, password=None,
useragent='Package Database Client/1.0', debug=None):
super(BaseClient, self).__init__(baseURL, username, password,
useragent, debug)
def collection_list(self):
'''Return a list of collections.'''
return client.send_request('/collection')
def package_owners(self, package, collectionName=None,
collectionVersion=None):
'''Return a mapping of release to owner for this package.'''
pkgData = client.send_request('/packages/name/%s' % (package),
{'collectionName': collectionName,
'collectionVersion': collectionVersion})
ownerMap = {}
for listing in pkgData['packageListings']:
ownerMap['-'.join(listing['collection']['name'],
listing['collection']['version'])] = \
listing['owneruser']
return ownerMap
A few things to note:
1. In our constructor we list a default baseURL and useragent. This is usually a good idea as we know
the URL of the Fedora Service we're connecting to and we want to know that people are using our
specific API.
2. Sometimes we'll want methods that are thin shells around the server methods like collection_list().
Other times we'll want to do more post processing to get specific results as package_owners() does.
Both types of methods are valid if they fit the needs of your API. If you find yourself writing more
of the latter, though, you may want to consider getting a new method implemented in the server that
can return results more appropriate to your needs as it could save processing on the server and
bandwidth downloading the data to get information that more closely matches what you need.
See pydoc fedora.accounts.fas2 for a module that implements a standard client API for the Fedora Account
System
Handling Errors
BaseClient will throw a variety of errors that can be caught to tell you what kind of error was
generated.
Exceptions
FedoraServiceError
The base of all exceptions raised by BaseClient. If your code needs to catch any of the listed
errors then you can catch that to do so.
ServerError
Raised if there's a problem communicating with the service. For instance, if we receive an HTML
response instead of JSON.
AuthError
If something happens during authentication, like an invalid usernsme or password, AuthError will
be raised. You can catch this to prompt the user for a new usernsme.
AppError
If there is a server side error when processing a request, the Fedora Service can alert the client
of this by setting certain flags in the response. BaseClient will see these flags and raise an
AppError. The name of the error will be stored in AppError's name field. The error's message
will be stored in message.
Example
Here's an example of the exceptions in action:
from fedora.client import ServerError, AuthError, AppError, BaseClient
import getpass
MAXRETRIES = 5
client = BaseClient('https://admin.fedoraproject.org/pkgdb')
for retry in range(0, MAXRETRIES):
try:
collectionData = client.send_request('/collections', auth=True)
except AuthError as e:
from six.moves import input
client.username = input('Username: ').strip()
client.password = getpass.getpass('Password: ')
continue
except ServerError as e:
print('Error talking to the server: %s' % e)
break
except AppError as e:
print('The server issued the following exception: %s: %s' % (
e.name, e.message))
for collection in collectionData['collections']:
print('%s %s' % (collection[0]['name'], collection[0]['version']))
OpenIdBaseClient
Applications that use OpenId to authenticate are not able to use the standard BaseClient because the
pattern of authenticating is very different. We've written a separate client object called
OpenIdBaseClient to do this.
EXISTING SERVICES
There are many Services in Fedora. Many of these have an interface that we can query and get back
information as JSON data. There is documentation here about both the services and the client modules
that can access them.
Fedora Account System
FAS is the Fedora Account System. It holds the account data for all of our contributors.
class fedora.client.AccountSystem(base_url='https://admin.fedoraproject.org/accounts/', *args, **kwargs)
An object for querying the Fedora Account System.
The Account System object provides a python API for talking to the Fedora Account System. It
abstracts the http requests, cookie handling, and other details so you can concentrate on the
methods that are important to your program.
WARNING:
If your code is trying to use the AccountSystem object to connect to fas for multiple users you
probably want to use FasProxyClient instead. If your code is trying to reuse a single instance
of AccountSystem for multiple users you definitely want to use FasProxyClient instead. Using
AccountSystem in these cases may result in a user being logged in as a different user. (This
may be the case even if you instantiate a new AccountSystem object for each user if
:attr:cache_session: is True since that creates a file on the file system that can end up
loading session credentials for the wrong person.
Changed in version 0.3.26: Added gravatar_url() that returns a url to a gravatar for a user.
Changed in version 0.3.33: Renamed gravatar_url() to avatar_url().
avatar_url(username, size=64, default=None, lookup_email=True, service=None)
Returns a URL to an avatar for a given username.
Avatars are drawn from third party services.
Parameters
• username -- FAS username to construct a avatar url for
• size -- size of the avatar. Allowed sizes are 32, 64, 140. Default: 64
• default -- If the service does not have a avatar image for the email address, this
url is returned instead. Default: the fedora logo at the specified size.
• lookup_email -- If true, use the email from FAS for gravatar.com lookups,
otherwise just append @fedoraproject.org to the username. For libravatar.org
lookups, this is ignored. The openid identifier of the user is used instead.
Note that gravatar.com lookups will be much slower if lookup_email is set to True
since we'd have to make a query against FAS itself.
• service -- One of 'libravatar' or 'gravatar'. Default: 'libravatar'.
Raises ValueError -- if the size parameter is not allowed or if the service is not one of
'libravatar' or 'gravatar'
Return type
str
Returns
url of a avatar for the user
If that user has no avatar entry, instruct the remote service to redirect us to the Fedora
logo.
If that user has no email attribute, then make a fake request to the third party service.
New in version 0.3.26.
create_group(name, display_name, owner, group_type, invite_only=0, needs_sponsor=0,
user_can_remove=1, prerequisite='', joinmsg='', apply_rules='None')
Creates a FAS group.
Parameters
• name -- The short group name (alphanumeric only).
• display_name -- A longer version of the group's name.
• owner -- The username of the FAS account which owns the new group.
• group_type -- The kind of group being created. Current valid options are git, svn,
hg, shell, and tracking.
• invite_only -- Users must be invited to the group, they cannot join on their own.
• needs_sponsor -- Users must be sponsored into the group.
• user_can_remove -- Users can remove themselves from the group.
• prerequisite -- Users must be in the given group (string) before they can join the
new group.
• joinmsg -- A message shown to users when they apply to the group.
• apply_rules -- Rules for applying to the group, shown to users before they apply.
Return type
munch.Munch
Returns
A Munch containing information about the group that was created.
New in version 0.3.29.
get_config(username, application, attribute)
Return the config entry for the key values.
Parameters
• username -- Username of the person
• application -- Application for which the config is set
• attribute -- Attribute key to lookup
Raises AppError -- if the server returns an exception
Returns
The unicode string that describes the value. If no entry matched the username,
application, and attribute then None is returned.
get_configs_like(username, application, pattern=u'*')
Return the config entries that match the keys and the pattern.
Note: authentication on the server will prevent anyone but the user or a fas admin from
viewing or changing their configs.
Parameters
• username -- Username of the person
• application -- Application for which the config is set
• pattern -- A pattern to select values for. This accepts * as a wildcard
character. Default='*'
Raises AppError -- if the server returns an exception
Returns
A dict mapping attribute to value.
gravatar_url(*args, **kwargs)
Deprecated - Use avatar_url.
New in version 0.3.26.
group_by_id(group_id)
Returns a group object based on its id
group_by_name(groupname)
Returns a group object based on its name
group_data(force_refresh=None)
Return administrators/sponsors/users and group type for all groups
Parameters
force_refresh -- If true, the returned data will be queried from the database, as
opposed to memcached.
Raises AppError -- if the query failed on the server
Returns
A dict mapping group names to the group type and the user IDs of the administrator,
sponsors, and users of the group.
New in version 0.3.8.
group_members(groupname)
Return a list of people approved for a group.
This method returns a list of people who are in the requested group. The people are all
approved in the group. Unapproved people are not shown. The format of data is:
\[{'username': 'person1', 'role_type': 'user'},
\{'username': 'person2', 'role_type': 'sponsor'}]
role_type can be one of 'user', 'sponsor', or 'administrator'.
New in version 0.3.2.
Changed in version 0.3.21: Return a Bunch instead of a DictContainer
insecure
If this attribute is set to True, do not check server certificates against their CA's.
This means that man-in-the-middle attacks are possible. You might turn this option on for
testing against a local version of a server with a self-signed certificate but it should be
off in production.
people_by_groupname(groupname)
Return a list of persons for the given groupname.
Parameters
groupname -- Name of the group to look up
Returns
A list of person objects from the group. If the group contains no entries, then an
empty list is returned.
people_by_id()
Deprecated Use people_by_key() instead.
Returns a dict relating user IDs to human_name, email, username, and bugzilla email
Changed in version 0.3.21: Return a Bunch instead of a DictContainer
people_by_key(key=u'username', search=u'*', fields=None)
Return a dict of people
For example:
>>> ret_val = FASCLIENT.people_by_key(
... key='email', search='toshio*', fields=['id'])
>>> ret_val.keys()
a.badger@[...].com
a.badger+test1@[...].com
a.badger+test2@[...].com
>>> ret_val.values()
100068
102023
102434
Parameters
• key -- Key used to organize the returned dictionary. Valid values are 'id',
'username', or 'email'. Default is 'username'.
• search -- Pattern to match usernames against. Defaults to the '*' wildcard which
matches everyone.
• fields --
Limit the data returned to a specific list of fields. The default is to retrieve
all fields. Valid fields are:
• affiliation
• alias_enabled
• bugzilla_email
• certificate_serial
• comments
• country_code
• creation
• email
• emailtoken
• facsimile
• gpg_keyid
• group_roles
• human_name
• id
• internal_comments
• ircnick
• last_seen
• latitude
• locale
• longitude
• memberships
• old_password
• password
• password_changed
• passwordtoken
• postal_address
• privacy
• roles
• ssh_key
• status
• status_change
• telephone
• timezone
• unverified_email
• username
Note that for most users who access this data, many of these fields will be set to
None due to security or privacy settings.
Returns
a dict relating the key value to the fields.
Changed in version 0.3.21: Return a Bunch instead of a DictContainer
Changed in version 0.3.26: Fixed to return a list with both people who have signed the CLA
and have not
people_query(constraints=None, columns=None)
Returns a list of dicts representing database rows
Parameters
• constraints -- A dictionary specifying WHERE constraints on columns
• columns -- A list of columns to be selected in the query
Raises AppError -- if the query failed on the server (most likely because the server was
given a bad query)
Returns
A list of dicts representing database rows (the keys of the dict are the columns
requested)
New in version 0.3.12.1.
person_by_id(person_id)
Returns a person object based on its id
person_by_username(username)
Returns a person object based on its username
proxy = None
set_config(username, application, attribute, value)
Set a config entry in FAS for the user.
Note: authentication on the server will prevent anyone but the user or a fas admin from
viewing or changing their configs.
Parameters
• username -- Username of the person
• application -- Application for which the config is set
• attribute -- The name of the config key that we're setting
• value -- The value to set this to
Raises AppError -- if the server returns an exception
user_data()
Return user data for all users in FAS
Note: If the user is not authorized to see password hashes, '*' is returned for the hash.
Raises AppError -- if the query failed on the server
Returns
A dict mapping user IDs to a username, password hash, SSH public key, email address,
and status.
New in version 0.3.8.
user_gencert()
Generate a cert for a user
user_id()
Returns a dict relating user IDs to usernames
verify_password(username, password)
Return whether the username and password pair are valid.
Parameters
• username -- username to try authenticating
• password -- password for the user
Returns
True if the username/password are valid. False otherwise.
Threadsafe Account System Access
It is not safe to use a single instance of the AccountSystem object in multiple threads. This is because
instance variables are used to hold some connection-specific information (for instance, the user who is
logging in). For this reason, we also provide the fedora.client.FasProxyClient object.
This is especially handy when writing authn and authz adaptors that talk to fas from a multithreaded
webserver.
class fedora.client.FasProxyClient(base_url='https://admin.fedoraproject.org/accounts/', *args, **kwargs)
A threadsafe client to the Fedora Account System.
get_user_info(auth_params)
Retrieve information about a logged in user.
Parameters
auth_params -- Auth information for a particular user. For instance, this can be a
username/password pair or a session_id. Refer to
fedora.client.proxyclient.ProxyClient.send_request() for all the legal values for
this.
Returns
a tuple of session_id and information about the user.
Raises AuthError -- if the auth_params do not give access
group_list(auth_params)
Retrieve a list of groups
Parameters
auth_params -- Auth information for a particular user. For instance, this can be a
username/password pair or a session_id. Refer to
fedora.client.proxyclient.ProxyClient.send_request() for all the legal values for
this.
Returns
a tuple of session_id and information about groups. The groups information is in
two fields:
groups contains information about each group
memberships
contains information about which users are members of which groups
Raises AuthError -- if the auth_params do not give access
login(username, password)
Login to the Account System
Parameters
• username -- username to send to FAS
• password -- Password to verify the username with
Returns
a tuple of the session id FAS has associated with the user and the user's account
information. This is similar to what is returned by
fedora.client.proxyclient.ProxyClient.get_user_info()
Raises AuthError -- if the username and password do not work
logout(session_id)
Logout of the Account System
Parameters
session_id -- a FAS session_id to remove from FAS
person_by_id(person_id, auth_params)
Retrieve information about a particular person
Parameters
auth_params -- Auth information for a particular user. For instance, this can be a
username/password pair or a session_id. Refer to
fedora.client.proxyclient.ProxyClient.send_request() for all the legal values for
this.
Returns
a tuple of session_id and information about the user.
Raises
• AppError -- if the server returns an exception
• AuthError -- if the auth_params do not give access
refresh_session(session_id)
Try to refresh a session_id to prevent it from timing out
Parameters
session_id -- FAS session_id to refresh
Returns
session_id that FAS has set now
verify_password(username, password)
Return whether the username and password pair are valid.
Parameters
• username -- username to try authenticating
• password -- password for the user
Returns
True if the username/password are valid. False otherwise.
verify_session(session_id)
Verify that a session is active.
Parameters
session_id -- session_id to verify is currently associated with a logged in user
Returns
True if the session_id is valid. False otherwise.
Bodhi, the Update Server
Bodhi is used to push updates from the build system to the download repositories. It lets packagers send
packages to the testing repository or to the update repository.
pythyon-fedora currently supports both the old Bodhi1 interface and the new Bodhi2 interface. By using
fedora.client.BodhiCLient, the correct one should be returned to you depending on what is running live on
Fedora Infrastructure servers.
FEDORA SERVICES
Authors
Toshio Kuratomi Luke Macken
Date 02 February 2009
For Version
0.3.x
In the loosest sense, a Fedora Service is a web application that sends data that BaseClient is able to
understand. This document defines things that a web application must currently do for BaseClient to
understand it.
TurboGears and fedora.tg
Functions and classes to help build a Fedora Service.
All current Fedora Services are written in TurboGears. Examples in this document will be for that
framework. However, other frameworks can be used to write a Service if they can correctly create the
data that BaseClient needs.
A Fedora Service differs from other web applications in that certain URLs for the web application are an
API layer that the Service can send and receive JSON data from for BaseClient to interpret. This imposes
certain constraints on what data can be sent and what data can be received from those URLs. The
fedora.tg module contains functions that help you write code that communicated well with BaseClient.
The TurboGears framework separates an application into model, view, and controller layers. The model is
typically a database and holds the raw data that the application needs. The view formats the data before
output. The controller makes decisions about what data to retrieve from the model and which view to pass
it onto. The code that you'll most often need to use from fedora.tg operates on the controller layer but
there's also code that works on the model and view behind the scenes.
Controllers
The controller is the code that processes an http request. It validates and processes requests and
parameters sent by the user, gets data from the model to satisfy the request, and then passes it onto a
view layer to be returned to the user. fedora.tg.utils contains several helpful functions for working
with controllers.
URLs as API
In TurboGears and most web frameworks, a URL is a kind of API for accessing the data your web app
provides. This data can be made available in multiple formats. TurboGears allows you to use one URL to
serve multiple formats by specifying a query parameter or a special header.
Selecting JSON Output
A URL in TurboGears can serve double duty by returning multiple formats depending on how it is called.
In most cases, the URL will return HTML or XHTML by default. By adding the query parameter,
tg_format=json you can switch from returning the default format to returning JSON data. You need to add
an @expose(allow_json=True) decorator [1] to your controller method to tell TurboGears that this
controller should return JSON data:
@expose(allow_json=True)
@expose(template='my.templates.amplifypage')
def amplify(self, data):
allow_json=True is a shortcut for this:
@expose("json", content_type="text/javascript",
as_format="json", accept_format="text/javascript")
That means that the controller method will use the json template (uses TurboJson to marshal the returned
data to JSON) to return data of type text/javascript when either of these conditions is met: a query
param of tg_format=json or an Accept: text/javascript header is sent.
BaseClient in python-fedora 0.1.x and 0.2.x use the query parameter method of selecting JSON output.
BaseClient 0.2.99.6 and 0.3.x use both the header method and the query parameter since the argument was
made that the header method is friendlier to other web frameworks. 0.4 intends to use the header alone.
If you use allow_json=True this change shouldn't matter. If you specify the @expose("json") decorator
only accept_format is set. So it is advisable to include both as_format and accept_format in your
decorator. Note that this is a limitation of TurboGears <= 1.0.4.4. If your application is only going
to run on TurboGears > 1.0.4.4 you should be able to use a plain @expose("json") [2].
[1] http://docs.turbogears.org/1.0/ExposeDecorator#same-method-different-template
[2] http://trac.turbogears.org/ticket/1459#comment:4
Why Two Formats from a Single URL?
When designing your URLs you might wonder why you'd want to return JSON and HTML from a single controller
method instead of having two separate controller methods. For instance, separating the URLs into their
own namespaces might seem logical: /app/json/get_user/USERNAME as opposed to /app/user/USERNAME. Doing
things with two URLs as opposed to one has both benefits and drawbacks.
Benefits of One Method Handling Multiple Formats
• Usually less code as there's only one controller method
• When a user sees a page that they want to get data from, they can get it as JSON instead of screen
scraping.
• Forces the application designer to think more about the API that is being provided to the users instead
of just the needs of the web page they are creating.
• Makes it easier to see what data an application will need to implement an alternate interface since you
can simply look at the template code to see what variables are being used on a particular page.
Benefits of Multiple Methods for Each Format
• Avoids special casing for error handlers (See below)
• Separates URLs that you intend users to grab JSON data from URLs where you only want to display HTML.
• Allows the URLs that support JSON to concentrate on trimming the size of the data sent while URLs that
only return HTML can return whole objects.
• Organization can be better if you don't have to include all of the pages that may only be useful for
user interface elements.
Personal use has found that allowing JSON requests on one controller method works well for cases where
you want the user to get data and for traditional form based user interaction. AJAX requests have been
better served via dedicated methods.
Return Values
The toplevel of the return values should be a dict. This is the natural return value for TurboGears
applications.
Marshaling
All data should be encoded in JSON before being returned. This is normally taken care of automatically
by TurboGears and simplejson. If you are returning non-builtin objects you may have to define an
__json__() method.
Unicode
simplejson (and probably other JSON libraries) will take care of encoding Unicode strings to JSON so be
sure that you are passing Unicode strings around rather than encoded byte strings.
Error Handling
In python, error conditions are handled by raising an exception. However, an exception object will not
propagate automatically through a return from the server. Instead we set several special variables in
the returned data to inform BaseClient of any errors.
At present, when BaseClient receives an error it raises an exception of its own with the exception
information from the server inside. Raising the same exception as the server is being investigated but
may pose security risks so hasn't yet been implemented.
exc
All URLs which return JSON data should set the exc variable when the method fails unexpectedly (a
database call failed, a place where you would normally raise an exception, or where you'd redirect to an
error page if a user was viewing the HTML version of the web app). exc should be set to the name of an
exception and tg_flash set to the message that would normally be given to the exception's constructor.
If the return is a success (expected values are being returned from the method or a value was updated
successfully) exc may either be unset or set to None.
tg_flash
When viewing the HTML web app, tg_flash can be set with a message to display to the user either on the
next page load or via an AJAX handler. When used in conjunction with JSON, exc=EXCEPTIONNAME, and
BaseClient, tg_flash should be set to an error message that the client can use to identify what went
wrong or display to the user. It's equivalent to the message you would normally give when raising an
exception.
Authentication Errors
Errors in authentication are a special case. Instead of returning an error with exc='AuthError' set, the
server should return with response.status = 403. BaseClient will see the 403 and raise an AuthError.
This is the signal for the client to ask the user for new credentials (usually a new username and
password).
NOTE:
Upstream TurboGears has switched to sending a 401 for authentication problems. However, their use of
401 is against the http specification (It doesn't set the 'WWW-Authentication' header) and it causes
problems for konqueror and webkit based browsers so we probably will not be switching.
Performing Different Actions when Returning JSON
So far we've run across three features of TurboGears that provide value to a web application but don't
work when returning JSON data. We provide a function that can code around this.
fedora.tg.utils.request_format() will return the format that the page is being returned as. Code can use
this to check whether JSON output is expected and do something different based on it:
output = {'tg_flash': 'An Error Occurred'}
if fedora.tg.utils.request_format() == 'json':
output['exc'] = 'ServerError'
else:
output['tg_template'] = 'my.templates.error'
return output
In this example, we return an error through our "exception" mechanism if we are returning JSON and return
an error page by resetting the template if not.
Redirects
Redirects do not play well with JSON [3] because TurboGears is unable to turn the function returned from
the redirect into a dictionary that can be turned into JSON.
Redirects are commonly used to express errors. This is actually better expressed using tg_template
because that method leaves the URL intact. That allows the end user to look for spelling mistakes in
their URL. If you need to use a redirect, the same recipe as above will allow you to split your code
paths.
[3] Last checked in TurboGears 1.0.4
tg_template
Setting what template is returned to a user by setting tg_template in the return dict (for instance, to
display an error page without changing the URL) is a perfectly valid way to use TurboGears.
Unfortunately, since JSON is simply another template in TurboGears you have to be sure not to interfere
with the generation of JSON data. You need to check whether JSON was requested using
fedora.tg.utils.request_format() and only return a different template if that's not the case. The recipe
above shows how to do this.
Validators
Validators are slightly different than the issues we've encountered so far. Validators are used to check
and convert parameters sent to a controller method so that only good data is dealt with in the controller
method itself. The problem is that when a validator detects a parameter that is invalid, it performs a
special internal redirect to a method that is its error_handler. We can't intercept this redirect
because it happens in the decorators before our method is invoked. So we have to deal with the aftermath
of the redirect in the error_handler method:
class NotNumberValidator(turbogears.validators.FancyValidator):
messages = {'Number': 'Numbers are not allowed'}
def to_python(self, value, state=None):
try:
number = turbogears.validators.Number(value.strip())
except:
return value
raise validators.Invalid(self.message('Number', state), value,
state)
class AmplifyForm(turbogears.widgets.Form):
template = my.templates.amplifyform
submit_text = 'Enter word to amplify'
fields = [
turbogears.widgets.TextField(name='data',
validator=NotNumberValidator())
]
amplify_form = AmplifyForm()
class mycontroller(RootController):
@expose(template='my.templates.errorpage', allow_json=True)
def no_numbers(self, data):
errors = fedora.tg.utils.jsonify_validation_errors()
if errors:
return errors
# Construct a dict to return the data error message as HTML via
# the errorpage template
pass
@validate(form=amplify_form)
@error_handler('no_numbers')
@expose(template='my.templates.amplifypage', allow_json=True)
def amplify(self, data):
return dict(data=data.upper())
When a user hits amplify()'s URL, the validator checks whether data is a number. If it is, it redirects
to the error_handler, no_numbers(). no_numbers() will normally return HTML which is fine if we're simply
hitting amplify() from a web browser. If we're hitting it from a BaseClient app, however, we need it to
return JSON data instead. To do that we use jsonify_validation_errors() which checks whether there was a
validation error and whether we need to return JSON data. If both of those are true, it returns a
dictionary with the validation errors. This dictionary is appropriate for returning from the controller
method in response to a JSON request.
WARNING:
When defining @error_handler() order of decorators can be important. The short story is to always
make @validate() and @error_handler() the first decorators of your method. The longer version is that
this is known to cause errors with the json request not being honored or skipping identity checks when
the method is its own error handler.
Expected Methods
Certain controller methods are necessary in order for BaseClient to properly talk to your service.
TurboGears can quickstart an application template for you that sets most of these variables correctly:
$ tg-admin quickstart -i -s -p my my
# edit my/my/controllers.py
login()
You need to have a login() method in your application's root. This method allows BaseClient to
authenticate against your Service:
@expose(template="my.templates.login")
+ @expose(allow_json=True)
def login(self, forward_url=None, previous_url=None, \*args, \**kw):
if not identity.current.anonymous \
and identity.was_login_attempted() \
and not identity.get_identity_errors():
+ # User is logged in
+ if 'json' == fedora.tg.utils.request_format():
+ return dict(user=identity.current.user)
+ if not forward_url:
+ forward_url = turbogears.url('/')
raise redirect(forward_url)
For non-TurboGears Implementors
If you are implementing a server in a non-TurboGears framework, note that one of the ways to reach the
login() method is through special parameters parsed by the TurboGears framework. BaseClient uses these
parameters instead of invoking the login() method directly as it saves a round trip when authenticating
to the server. It will be necessary for you to implement handling of these parameters (passed via POST)
on your application as well.
The parameters are: user_name, password, and login. When these three parameters are sent to the server,
the server authenticates the user and records their information before deciding what information to
return to them from the URL.
logout()
The logout() method is similar to login(). It also needs to be modified to allow people to connect to it
via JSON:
- @expose()
+ @expose(allow_json=True)
def logout(self):
identity.current.logout()
+ if 'json' in fedora.tg.utils.request_format():
+ return dict()
raise redirect("/")
CSRF Protection
For an overview of CSRF and how to protect TurboGears 1 based services, look at this document.
CSRF Protection
Authors
Toshio Kuratomi
Date 21 February 2009
For Version
0.3.x
CSRF, Cross-Site Request Forgery is a technique where a malicious website can gain access to a Fedora
Service by hijacking a currently open session in a user's web browser. This technique affects
identification via SSL Certificates, cookies, or anything else that the browser sends to the server
automatically when a request to that server is made. It can take place over both GET and POST. GET
requests just need to hit a vulnerable URL. POST requests require JavaScript to construct the form that
is sent to the vulnerable server.
NOTE:
If you just want to implement this in Fedora-Services, skip to the Summary of Changes Per App section
How CSRF Works
1. A vulnerable web service allows users to change things on their site using just a cookie for
authentication and submission of a form or by hitting a URL with an img tag.
2. A malicious website is crafted that looks harmless but has JavaScript or an img tag that sends a
request to the web service with the form data or just hits the vulnerable URL.
3. The user goes somewhere that people who are frequently logged into the site are at and posts something
innocuous that gets people to click on the link to the malicious website.
4. When a user who is logged into the vulnerable website clicks on the link, their web browser loads the
page. It sees the img tag and contacts the vulnerable website to request the listed URL sending the
user's authentication cookie automatically.
5. The vulnerable server performs the action that the URL requires as the user whose cookies were sent
and the damage is done... typically without the user knowing any better until much later.
How to Prevent It
Theory
In order to remove this problem we need to have a shared secret between the user's browser and the web
site that is only available via the http request-response. This secret is required in order for any
actions to be performed on the site. Because the Same Origin Policy prevents the malicious website from
reading the web page itself, a shared secret passed in this manner will prevent the malicious site from
performing any actions. Note that this secret cannot be transferred from the user's browser to the
server via a cookie because this is something that the browser does automatically. It can, however, be
transferred from the server to the browser via a cookie because the browser prevents scripts from other
domains from reading the cookies.
Practice
The strategy we've adopted is sometimes called double submit. Every time we POST a form or make a GET
request that requires us to be authenticated we must also submit a token consisting of a hash of the
tg-visit to show the server that we were able to read either the cookie or the response from a previous
request. We store the token value in a GET or POST parameter named _csrf_token. If the server receives
the tg-visit without the _csrf_token parameter, the server renders the user anonymous until the user
clicks another link.
NOTE:
We hash the tg-visit session to make the token because we sometimes send the token as a parameter in
GET requests so it will show up in the servers http logs.
Verifying the Token
The jsonfasprovider1 does the work of verifying that _csrf_token has been set and that it is a valid hash
of the tg-visit token belonging to the user. The sequence of events to verify a user's identity follows
this outline:
1. If username and password given
1. Verify with the identity provider that the username and password match
1. [YES] authenticate the user
2. [NO] user is anonymous.
2. if tg-visit cookie present
1. if session_id from tg-visit is in the db and not expired and (sha1sum(tg-visit) matches _csrf_token
given as a (POST variable or GET variable)
1. [YES] authenticate the user
2. [NO] Take user to login page that just requires the user to click a link. Clicking the link
shows that it's not just JavaScript in the browser attempting to load the page but an actual
user. Once the link is clicked, the user will have a _csrf_token. If the link is not clicked
the user is treated as anonymous.
2. Verify via SSL Certificate
1. SSL Certificate is not revoked and able to retrieve info for the username and
(sha1sum(tg-visit) matches _csrf_token given as a POST variable or GET variable)
1. [YES] authenticate the user
2. [NO] Take user to login page that just requires the user to click a link. Clicking the link
shows that it's not just JavaScript in the browser attempting to load the page but an actual
user. Once the link is clicked, the user will have a _csrf_token. If the link is not
clicked the user is treated as anonymous.
This work should mostly happen behind the scenes so the application programmer does not need to worry
about this.
SEE ALSO:
The jsonfasprovider1 documentation has more information on methods that are provided by the identity
provider in case you do need to tell what the authentication token is and whether it is missing.
Getting the Token into the Page
Embedding the CSRF token into the URLs that the user can click on is the other half of this problem.
fedora.tg.utils provides two functions to make this easier.
This function does everything tg.url() does in the templates. In addition it makes sure that _csrf_token
is appended to the URL.
This function sets config values to allow _csrf_token to be passed to any URL on the server and makes
turbogears.url() point to our url() function. Once this is run, the tg.url() function you use in the
templates will make any links that use with it contain the CSRF protecting token.
Intra-Application Links
We support single sign-on among the web applications we've written for Fedora. In order for this to
continue working with this CSRF protection scheme we have to add the _csrf_token parameter to links
between Fedora-Services. Linking from the PackageDB to Bodhi, for instance, needs to have the hash
appended to the URL for Bodhi. Our url() function is capable of generating these by passing the full URL
into the function like this:
<a href="${tg.url('https://admin.fedoraproject.org/updates')}">Bodhi</a>
NOTE:
You probably already use tg.url() for links within you web app to support server.webpath. Adding
tg.url() to external links is new. You will likely have to update your application's templates to
call url() on these URLs.
Logging In
Each app's login() controller method and templates need to be modified in several ways.
Templates
For the templates, python-fedora provides a set of standard templates that can be used to add the token.
Genshi version of templates to make adding certain Fedora widgets easier.
fedora.tg.templates.genshi.login.html
Module author: Toshio Kuratomi <tkuratom@redhat.com>
New in version 0.3.10.
Include this using::
<xi:include href="${tg.fedora_template('login.html')}" />
fedora.tg.templates.genshi.login.html.loginform([message])
message
Any text or elements contained by the <loginform> tag will be shown as a message to the user.
This is generally used to show status of the last login attempt ("Please provide your
credentials", "Supplied credentials were not correct", etc)
A match template for the main login form. This is a CSRF token-aware login form that will prompt for
username and password when no session identity is present and ask the user to click a link if they merely
lack a token.
Typical usage would be:
<loginform>${message}</loginform>
fedora.tg.templates.genshi.login.html.logintoolitem(@href=URL)
@href If an href attribute is present for this tag, when a user is logged in, their username or
display_name will be a link to the URL.
A match template to add an entry to a toolbar. The entry will contain the user's username and a logout
button if the user is logged in, a verify login button if the user has a session cookie but not a CSRF
token, or a login button if the user doesn't have a session cookie.
Typical usage looks like this:
<ul class="toolbar" id="#main-toolbar">
<logintoolitem href="${tg.url('/users/info')}" />
</ul>
fedora.tg.templates.genshi.jsglobals.html
Module author: Toshio Kuratomi <tkuratom@redhat.com>
New in version 0.3.10.
Include this using::
<xi:include href="${tg.fedora_template('jsglobals.html')}" />
fedora.tg.templates.genshi.jsglobals.html.jsglobals()
A match template to add global variables to a page. Typically, you'd include this in your master.html
template and let it be added to every other page from there. This adds the following variables in the
fedora namespace for other scripts to access:
fedora.baseurl
URL fragment to prepend to any calls to the application. In a TurboGears application, this is
the scheme, host, and server.webpath. Example: https://admin.fedoraproject.org/pkgdb/. This
may be a relative link.
fedora.identity.anonymous
If true, there will be no other variables in the fedora.identity namespace. If false, these
variables are defined:
fedora.identity.userid
Numeric, unique identifier for the user
fedora.identity.username
Publically visible unique identifier for the user
fedora.identity.display_name
Common human name for the user
fedora.identity.token
csrf token for this user's session to be added to urls that query the server.
Typical usage would be:
<jsglobals />
Using the <loginform> template will give you a login form that automatically does several things for you.
1. The forward_url and previous_url parameters that are passed in hidden form elements will be run
through tg.url() in order to get the _csrf_token added.
2. The page will allow "Click through validation" of a user when they have a valid tg-visit but do not
have a _csrf_token.
Here's a complete login.html from the pkgdb to show what this could look like:
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN"
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html xmlns="http://www.w3.org/1999/xhtml"
xmlns:py="http://genshi.edgewall.org/"
xmlns:xi="http://www.w3.org/2001/XInclude">
<xi:include href="master.html" />
<xi:include href="${tg.fedora_template('login.html')}" />
<head>
<meta content="text/html; charset=UTF-8"
http-equiv="content-type" py:replace="''"/>
<title>Login to the PackageDB</title>
</head>
<body>
<loginform>${message}</loginform>
</body>
</html>
You should notice that this looks like a typical genshi template in your project with two new features.
The <loginform> tag in the body that's defined in fedora.tg.templates.genshi is used in the body to pull
in the login formand the <xi:include> of login.html uses tg.fedora_template() to load the template from
python-fedora. This function resides in fedora.tg.utils and is added to the tg template variable when
enable_csrf() is called at startup. It does the following:
The second match template in login.html is to help you modify the login and logout links that appear at
the top of a typical application's page. This is an optional change that lets the links display a
click-through login link in addition to the usual login and logout. To use this, you would follow the
example to add a toolbar with the <logintoolitem> into your master template. Here's some snippets from a
master.html to illustrate:
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN"
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html xmlns="http://www.w3.org/1999/xhtml"
xmlns:py="http://genshi.edgewall.org/"
xmlns:xi="http://www.w3.org/2001/XInclude">
[...]
<body py:match="body" py:attrs="select('@*')">
[...]
<div id="head">
<h1><a href="http://fedoraproject.org/index.html">Fedora</a></h1>
<ul class="toolbar" id="#main-toolbar">
<logintoolitem href="${tg.url('/users/info')}" />
</ul>
</div>
[...]
</body>
<xi:include href="${tg.fedora_template('login.html')}" />
<xi:include href="${tg.fedora_template('jsglobals.html')}" />
</html>
WARNING:
Notice that the <xi:include> of login.html happens after the <body> tag? It is important to do that
because the <body> tag in master.html is a match template just like <logintoolitem>. In genshi, the
order in which match templates are defined is significant.
If you need to look at these templates to modify them yourself (perhaps to port them to a different
templating language) you can find them in fedora/tg/templates/genshi/login.html in the source tarball.
Controllers
Calling Fedora-Services from JavaScript poses one further problem. Sometimes the user will not have a
current CSRF token and will need to log in. When this is done with a username and password there's no
problem because the username and password are sufficient to prove the user is in control of the browser.
However, when using an SSL Certificate, we need the browser to log in and then use the new CSRF token to
show the user is in control. Since JavaScript calls are most likely going to request the information as
JSON, we need to provide the token in the dict returned from login(). You can either modify your login()
method to do that or use the one provided in fedora.tg.controllers.login().
NOTE:
Fedora Services do not currently use certificate authentication but doing it would not take much code.
Working out the security ramifications within Infrastructure is the main sticking point to turning
this on.
NOTE:
The Fedora Account System has a slightly different login() method. This is because it has to handle
account expiration, mandatory password resets, and other things tied to the status of an account.
Other Fedora Services rely on FAS to do this instead of having to handle it themselves.
AJAX
Making JavaScript calls requires that we can get the token and add it to the URLs we request. Since
JavaScript doesn't have a standard library of crypto functions we provide the token by setting it via a
template so we do not have to calculate it on the client. This has the added bonus of propagating a
correct token even if we change the hash function later. Making use of the token can then be done in ad
hoc JavaScript code but is better handled via a dedicated JavaScript method like the
fedora.dojo.BaseClient.
Template
WARNING:
Just like login.html, the <xi:include> tag needs to come after the <body> tag since they're both match
templates.
JavaScript
The fedora.dojo.BaseClient class has been modified to send and update the csrf token that is provided by
fedora.identity.token. It is highly recommended that you use this library or another like it to make all
calls to the server. This keeps you from having to deal with adding the CSRF token to every call
yourself.
Here's a small bit of sample code:
<script type="text/javascript" src="/js/dojo/dojo.js"></script>
<script type="text/javascript">
dojo.require('fedora.dojo.BaseClient');
dojo.addOnLoad(function() {
pkgdb = new fedora.dojo.BaseClient(fedora.baseurl, {
username: '', password:''});
pkg_data = pkgdb.start_request('/packages/name/python',
{req_params: {collectionName: 'Fedora',
collectionVersion: 'devel'}});
});
</script>
Summary of Changes Per App
• On startup, run enable_csrf(). This could be done in a start-APP.py and APP.wsgi scripts. Code
like this will do it:
from turbogears import startup
from fedora.tg.utils import enable_csrf
startup.call_on_startup.append(enable_csrf)
• Links to other Fedora-Services in the templates must be run through the tg.url() method so that the
CSRF token can be appended.
• You must use an updated login template. Using the one provided in python-fedora is possible by
changing your login template as shown in the Templates section.
• Optional: update the master template that shows the Login/Logout Link to have a "Verify Login"
button. You can add one to a toolbar in your application following the instructions in the
Templates section.
• Use an updated identity provider from python-fedora. At this time, you need python-fedora 0.3.10 or
later which has a jsonfasprovider2 and jsonfasvisit2 that provide CSRF protection. The original
jsonfasprovider1 is provided for applications that have not yet started using enable_csrf() so you
have to make this change in your configuration file (APPNAME/config/app.cfg)
• Get the CSRF token into your forms and URLs. The recommended way to do this is to use tg.url() in
your forms for URLs that are local to the app or are for Fedora-Services.
• Update your login() method to make sure you're setting forward_url = request.path_info rather than
request.path. One easy way to do this is to use the login() and logout() as documented in
Controllers
• Add the token and other identity information so JavaScript can get at it. Use the jsglobals
template to accomplish this.
This one still needs to be implemented
• AJAX calls need to be enhanced to append the CSRF token to the data. This is best done using a
JavaScript function for this like the fedora.dojo.BaseClient library.
Using SABase
fedora.tg.json contains several functions that help to convert SQLAlchemy objects into JSON. For the
most part, these do their work behind the scenes. The SABase object, however, is one that you might need
to take an active role in using.
When you return an SQLAlchemy object in a controller to a template, the template is able to access any of
the relations mapped to it. So, instead of having to construct a list of people records from a table and
the the list of groups that each of them are in you can pass in the list of people and let your template
reference the relation properties to get the groups. This is extremely convenient for templates but has
a negative effect when returning JSON. Namely, the default methods for marshaling SQLAlchemy objects to
JSON only return the attributes of the object, not the relations that are linked to it. So you can
easily run into a situation where someone querying the JSON data for a page will not have all the
information that a template has access to.
SABase fixes this by allowing you to specify relations that your SQLAlchemy backed objects should marshal
as JSON data.
Further information on SABase can be found in the API documentation:
pydoc fedora.tg.json
Example
SABase is a base class that you can use when defining objects in your project's model. So the first step
is defining the classes in your model to inherit from SABase:
from fedora.tg.json import SABase
from sqlalchemy import Table, Column, Integer, String, MetaData, ForeignKey
from turbogears.database import metadata, mapper
class Person(SABase):
pass
PersonTable = Table('person', metadata
Column('name', String, primary_key=True),
)
class Address(SABase):
pass
AddressTable = Table (
Column('id', Integer, primary_key=True),
Column('street', string),
Column('person_id', Integer, ForeignKey('person.name')
)
mapper(PersonTable, Person)
mapper(AddressTable, Address, properties = {
person: relation(Person, backref = 'addresses'),
})
The next step is to tell SABase which properties should be copied (this allows you to omit large trees of
objects when you only need the data from a few of them):
@expose('my.templates.about_me')
@expose(allow_json=True)
def my_info(self):
person = Person.query.filter_by(name='Myself').one()
person.jsonProps = {'Person': ['addresses']}
return dict(myself=person}
Now, when someone requests JSON data from my_info, they should get back a record for person that includes
a property addresses. Addresses will be a list of address records associated with the person.
How it Works
SABase adds a special __json__() method to the class. By default, this method returns a dict with all of
the attributes that are backed by fields in the database.
Adding entries to jsonProps adds the values for those properties to the returned dict as well. If you
need to override the __json__() method in your class you probably want to call SABase's __json__() unless
you know that neither you nor any future subclasses will need it.
Using __json__()
Sometimes you need to return an object that isn't a basic python type (list, tuple, dict, number. string,
etc). When that occurs, simplejson won't know how to marshal the data into JSON until you write own
method to transform the values. If this method is named __json__(), TurboGears will automatically
perform the conversion when you return the object.
Example:
class MyObject(object):
def _init__(self, number):
self.someNumber = number
self.cached = None
def _calc_data(self):
if not self.cached:
self.cached = self.someNumber * 2
return self.cached
twiceData = property(_calc_data)
def __json__(self):
return {'someNumber': self.someNumber, 'twiceData': self.twiceData}
In this class, you have a variable and a property. If you were to return it from a controller method
without defining the __json__() method, TurboGears would give you an error that it was unable to adapt
the object to JSON. The JSON method transforms the object into a dict with sensibly named values for the
variable and property so that simplejson is able to marshal the data to JSON. Note that you will often
have to choose between space (more data takes more bandwidth to deliver to the end user) and completeness
(you need to return enough data so the client isn't looking for another method that can complete its
needs) when returning data.
AUTHENTICATION TO FAS
The Fedora-Account-System has a JSON interface that we make use of to authenticate users in our web apps.
Currently, there are two modes of operation. Some web apps have single sign-on capability with FAS.
These are the TurboGears applications that use the jsonfasprovider. Other apps do not have single
sign-on but they do connect to FAS to verify the username and password so changing the password in FAS
changes it everywhere.
TurboGears Identity Provider 2
An identity provider with CSRF protection.
This will install as a TurboGears identity plugin. To use it, set the following in your
APPNAME/config/app.cfg file:
identity.provider='jsonfas2'
visit.manager='jsonfas2'
SEE ALSO:
CSRF-Protection
Turbogears Identity Provider 1
These methods are deprecated because they do not provide the CSRF protection of TurboGears Identity
Provider 2. Please use that identity provider instead.
Django Authentication Backend
Fedora Django Authentication Backend
Authors
Ignacio Vazquez-Abrams
Date 23 Feb 2009
For Version
0.3.x
The django.auth package provides an authentication backend for Django projects.
NOTE:
Django authentication does not provide single sign-on with other Fedora web apps. it also does not
provide CSRF protection. Look at the way that Dango's builtin forms implement CSRF protection for
guidance on how to protect against this sort of attack.
fedora.django.auth
As FAS users are authenticated they are added to FasUser. FAS groups are added to Group both during
syncdb and when a user is authenticated.
Integrating into a Django Project
Add the following lines to the project's settings.py:
AUTHENTICATION_BACKENDS = (
'fedora.django.auth.backends.FasBackend',
)
FAS_USERNAME = '<username>'
FAS_PASSWORD = '<password>'
FAS_USERAGENT = '<user agent>'
FAS_URL = '<base URL of FAS server>'
FAS_ADMINS = ( ... )
FAS_USERNAME and FAS_PASSWORD are used to retrieve group information during syncdb as well as to retrieve
users via the authentication backend. They should be set to a low-privilege account that can read group
and user information.
FAS_USERAGENT is the string used to identify yourself to the FAS server.
FAS_URL is the base URL of the FAS server to authenticate against.
FAS_ADMINS is a tuple of usernames that you want to have superuser rights in the Django project.
Add fedora.django.auth.middleware.FasMiddleware to the MIDDLEWARE_CLASSES tuple, between
django.contrib.sessions.middleware.SessionMiddleware and
django.contrib.auth.middleware.AuthenticationMiddleware.
Additionally, set FAS_GENERICEMAIL to False in order to use the email address specified in FAS instead of
<username>``@fedoraproject.org``.
Add fedora.django.auth to INSTALLED_APPS.
Finally, run python manage.py syncdb to add the models for the added app to the database.
WARNING:
The User.first_name and User.last_name attributes are always empty since FAS does not have any
equivalents. The name read-only property results in a round trip to the FAS server.
Flask Auth Plugin
FAS Flask Auth Plugin
Authors
Toshio Kuratomi, Ian Weller
Date 29 October 2012
For Version
0.3.x
The Fedora-Account-System has a JSON interface that we make use of to authenticate users in our web apps.
For our Flask applications we have an identity provider that has single sign-on with our TurboGears 1 and
2 applications. It does not protect against CSRF attacks in the identity layer. The flask-wtf forms
package should be used to provide that.
Configuration
The FAS auth plugin has several config values that can be used to control how the auth plugin functions.
You can set these in your application's config file.
FAS_BASE_URL
Set this to the URL of the FAS server you are authenticating against. Default is "‐
https://admin.fedoraproject.org/accounts/"
FAS_USER_AGENT
User agent string to be used when connecting to FAS. You can set this to something specific to
your application to aid in debugging a connection to the FAS server as it will show up in the FAS
server's logs. Default is "Flask-FAS/0.3"
FAS_CHECK_CERT
When set, this will check the SSL Certificate for the FAS server to make sure that it is who it
claims to be. This is useful to set to False when testing against a local FAS server but should
always be set to True in production. Default: True
FAS_COOKIE_NAME
The name of the cookie used to store the session id across the Fedora Applications that support
single sign-on. Default: "tg-visit"
FAS_FLASK_COOKIE_REQUIRES_HTTPS
When this is set to True, the session cookie will only be returned to the server via ssl (https).
If you connect to the server via plain http, the cookie will not be sent. This prevents sniffing
of the cookie contents. This may be set to False when testing your application but should always
be set to True in production. Default is True.
Sample Application
The following is a sample, minimal flask application that uses fas_flask for authentication:
#!/usr/bin/python -tt
# Flask-FAS - A Flask extension for authorizing users with FAS
# Primary maintainer: Ian Weller <ianweller@fedoraproject.org>
#
# Copyright (c) 2012, Red Hat, Inc.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
#
# * Redistributions of source code must retain the above copyright notice, this
# list of conditions and the following disclaimer.
# * Redistributions in binary form must reproduce the above copyright notice,
# this list of conditions and the following disclaimer in the documentation
# and/or other materials provided with the distribution.
# * Neither the name of the Red Hat, Inc. nor the names of its contributors may
# be used to endorse or promote products derived from this software without
# specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ''AS IS'' AND ANY
# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
# DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
# DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
# (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
# ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
# This is a sample application. In addition to using Flask-FAS, it uses
# Flask-WTF (WTForms) to handle the login form. Use of Flask-WTF is highly
# recommended because of its CSRF checking.
import flask
from flask.ext import wtf
from flask.ext.fas import FAS, fas_login_required
# Set up Flask application
app = flask.Flask(__name__)
# Set up FAS extension
fas = FAS(app)
# Application configuration
# SECRET_KEY is necessary to CSRF in WTForms. It needs to be secret to
# make the csrf tokens unguessable but if you have multiple servers behind
# a load balancer, the key needs to be the same on each.
app.config['SECRET_KEY'] = 'change me!'
# Other configuration options for Flask-FAS:
# FAS_BASE_URL: the base URL for the accounts system
# (default https://admin.fedoraproject.org/accounts/)
# FAS_CHECK_CERT: check the SSL certificate of FAS (default True)
# FAS_FLASK_COOKIE_REQUIRES_HTTPS: send the 'secure' option with
# the login cookie (default True)
# You should use these options' defaults for production applications!
app.config['FAS_BASE_URL'] = 'https://fakefas.fedoraproject.org/accounts/'
app.config['FAS_CHECK_CERT'] = False
app.config['FAS_FLASK_COOKIE_REQUIRES_HTTPS'] = False
# A basic login form
class LoginForm(wtf.Form):
username = wtf.TextField('Username', [wtf.validators.Required()])
password = wtf.PasswordField('Password', [wtf.validators.Required()])
# Inline templates keep this test application all in one file. Don't do this in
# a real application. Please.
TEMPLATE_START = """
<h1>Flask-FAS test app</h1>
{% if g.fas_user %}
<p>Hello, {{ g.fas_user.username }} —
<a href="{{ url_for("logout") }}">Log out</a>
{% else %}
<p>You are not logged in —
<a href="{{ url_for("auth_login", next=request.url) + '' }}">Log in</a>
{% endif %}
— <a href="{{ url_for("index") }}">Main page</a></p>
"""
@app.route('/')
def index():
data = TEMPLATE_START
data += '<p><a href="%s">Check if you are cla+1</a></p>' % \
flask.url_for('claplusone')
data += '<p><a href="%s">See a secret message (requires login)</a></p>' % \
flask.url_for('secret')
return flask.render_template_string(data)
@app.route('/login', methods=['GET', 'POST'])
def auth_login():
# Your application should probably do some checking to make sure the URL
# given in the next request argument is sane. (For example, having next set
# to the login page will cause a redirect loop.) Some more information:
# http://flask.pocoo.org/snippets/62/
if 'next' in flask.request.args:
next_url = flask.request.args['next']
else:
next_url = flask.url_for('index')
# If user is already logged in, return them to where they were last
if flask.g.fas_user:
return flask.redirect(next_url)
# Init login form
form = LoginForm()
# Init template
data = TEMPLATE_START
data += ('<p>Log into the <a href="{{ config.FAS_BASE_URL }}">'
'Fedora Accounts System</a>:')
# If this is POST, process the form
if form.validate_on_submit():
if fas.login(form.username.data, form.password.data):
# Login successful, return
return flask.redirect(next_url)
else:
# Login unsuccessful
data += '<p style="color:red">Invalid login</p>'
data += """
<form action="" method="POST">
{% for field in [form.username, form.password] %}
<p>{{ field.label }}: {{ field|safe }}</p>
{% if field.errors %}
<ul style="color:red">
{% for error in field.errors %}
<li>{{ error }}</li>
{% endfor %}
</ul>
{% endif %}
{% endfor %}
<input type="submit" value="Log in">
{{ form.csrf_token }}
</form>"""
return flask.render_template_string(data, form=form)
@app.route('/logout')
def logout():
if flask.g.fas_user:
fas.logout()
return flask.redirect(flask.url_for('index'))
# This demonstrates the use of the fas_login_required decorator. The
# secret message can only be viewed by those who are logged in.
@app.route('/secret')
@fas_login_required
def secret():
data = TEMPLATE_START + '<p>Be sure to drink your Ovaltine</p>'
return flask.render_template_string(data)
# This demonstrates checking for group membership inside of a function.
# The flask_fas adapter also provides a cla_plus_one_required decorator that
# can restrict a url so that you can only access it from an account that has
# cla +1.
@app.route('/claplusone')
def claplusone():
data = TEMPLATE_START
if not flask.g.fas_user:
# Not logged in
return flask.render_template_string(data +
'<p>You must log in to check your cla +1 status</p>')
non_cla_groups = [x.name for x in flask.g.fas_user.approved_memberships
if x.group_type != 'cla']
if len(non_cla_groups) > 0:
data += '<p>Your account is cla+1.</p>'
else:
data += '<p>Your account is <em>not</em> cla+1.</p>'
return flask.render_template_string(data)
if __name__ == '__main__':
app.run(debug=True)
Flask FAS OpenId Auth Plugin
The flask_openid provider is an alternative to the flask_fas auth plugin. It leverages our FAS-OpenID
server to do authn and authz (group memberships). Note that not every feature is available with a
generic OpenID provider -- the plugin depends on the OpenID provider having certain extensions in order
to provide more than basic OpenID auth.
• Any compliant OpenID server should allow you to use the basic authn features of OpenID OpenID
authentication core: http://openid.net/specs/openid-authentication-2_0.html
• Retrieving simple information about the user such as username, human name, email is done with sreg:
http://openid.net/specs/openid-simple-registration-extension-1_0.html which is an extension supported
by many providers.
• Advanced security features such as requiring a user to re-login to the OpenID provider or specifying
that the user login with a hardware token requires the PAPE extension:
http://openid.net/specs/openid-provider-authentication-policy-extension-1_0.html
• To get groups information, the provider must implement the https://dev.launchpad.net/OpenIDTeams
extension.
• We have extended the teams extension so you can request a team name of _FAS_ALL_GROUPS_ to retrieve
all the groups that a user belongs to. Without this addition to the teams extension you will need to
manually configure which groups you are interested in knowing about. See the documentation for how
to do so.
• Retrieving information about whether a user has signed a CLA (For Fedora, this is the Fedora Project
Contributor Agreement). http://fedoraproject.org/specs/open_id/cla
If the provider you use does not support one of these extensions, the plugin should still work but
naturally, it will return empty values for the information that the extension would have provided.
FAS Flask OpenID Auth Plugin
Authors
Patrick Uiterwjk
Date 18 February 2013
For Version
0.3.x
The Fedora-Account-System has a OpenID provider that applications can use to authenticate users in web
apps. For our Flask applications we have an identity provider that uses this OpenID service to
authenticate users. It is almost completely compatible with flask_fas except that it does not use the
username/password provided by the client application (it is silently ignored). It can be configured to
use any OpenID authentication service that implements the OpenID Teams Extension, Simple Registration
Extension and CLA Extension.
Configuration
The FAS OpenID auth plugin has several config values that can be used to control how the auth plugin
functions. You can set these in your application's config file.
FAS_OPENID_ENDPOINT
Set this to the OpenID endpoint url you are authenticating against. Default is "‐
http://id.fedoraproject.org/"
FAS_CHECK_CERT
When set, this will check the SSL Certificate for the FAS server to make sure that it is who it
claims to be. This is useful to set to False when testing against a local FAS server but should
always be set to True in production. Default: True
Sample Application
The following is a sample, minimal flask application that uses fas_flask for authentication:
#!/usr/bin/python -tt
# Flask-FAS-OpenID - A Flask extension for authorizing users with OpenID
# Primary maintainer: Patrick Uiterwijk <puiterwijk@fedoraproject.org>
#
# Copyright (c) 2012-2013, Red Hat, Inc., Patrick Uiterwijk
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
#
# * Redistributions of source code must retain the above copyright notice, this
# list of conditions and the following disclaimer.
# * Redistributions in binary form must reproduce the above copyright notice,
# this list of conditions and the following disclaimer in the documentation
# and/or other materials provided with the distribution.
# * Neither the name of the Red Hat, Inc. nor the names of its contributors may
# be used to endorse or promote products derived from this software without
# specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ''AS IS'' AND ANY
# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
# DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
# DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
# (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
# ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
# This is a sample application.
import flask
from flask_fas_openid import fas_login_required, cla_plus_one_required, FAS
# Set up Flask application
app = flask.Flask(__name__)
# Set up FAS extension
fas = FAS(app)
# Application configuration
# SECRET_KEY is necessary for the Flask session system. It needs to be secret to
# make the sessions secret but if you have multiple servers behind
# a load balancer, the key needs to be the same on each.
app.config['SECRET_KEY'] = 'change me!'
# Other configuration options for Flask-FAS-OpenID:
# FAS_OPENID_ENDPOINT: the OpenID endpoint URL
# (default http://id.fedoraproject.org/)
# FAS_CHECK_CERT: check the SSL certificate of FAS (default True)
# You should use these options' defaults for production applications!
app.config['FAS_OPENID_ENDPOINT'] = 'http://id.fedoraproject.org/'
app.config['FAS_CHECK_CERT'] = True
# Inline templates keep this test application all in one file. Don't do this in
# a real application. Please.
TEMPLATE_START = """
<h1>Flask-FAS-OpenID test app</h1>
{% if g.fas_user %}
<p>Hello, {{ g.fas_user.username }} —
<a href="{{ url_for("logout") }}">Log out</a>
{% else %}
<p>You are not logged in —
<a href="{{ url_for("auth_login", next=request.url) + '' }}">Log in</a>
{% endif %}
— <a href="{{ url_for("index") }}">Main page</a></p>
"""
@app.route('/')
def index():
data = TEMPLATE_START
data += '<p><a href="%s">Check if you are cla+1</a></p>' % \
flask.url_for('claplusone')
data += '<p><a href="%s">See a secret message (requires login)</a></p>' % \
flask.url_for('secret')
return flask.render_template_string(data)
@app.route('/login', methods=['GET', 'POST'])
def auth_login():
# Your application should probably do some checking to make sure the URL
# given in the next request argument is sane. (For example, having next set
# to the login page will cause a redirect loop.) Some more information:
# http://flask.pocoo.org/snippets/62/
if 'next' in flask.request.args:
next_url = flask.request.args['next']
else:
next_url = flask.url_for('index')
# If user is already logged in, return them to where they were last
if flask.g.fas_user:
return flask.redirect(next_url)
return fas.login(return_url=next_url)
@app.route('/logout')
def logout():
if flask.g.fas_user:
fas.logout()
return flask.redirect(flask.url_for('index'))
# This demonstrates the use of the fas_login_required decorator. The
# secret message can only be viewed by those who are logged in.
@app.route('/secret')
@fas_login_required
def secret():
data = TEMPLATE_START + '<p>Be sure to drink your Ovaltine</p>'
return flask.render_template_string(data)
# This demonstrates checking for group membership inside of a function.
# The flask_fas adapter also provides a cla_plus_one_required decorator that
# can restrict a url so that you can only access it from an account that has
# cla +1.
@app.route('/claplusone')
@cla_plus_one_required
def claplusone():
data = TEMPLATE_START
data += '<p>Your account is cla+1.</p>'
return flask.render_template_string(data)
if __name__ == '__main__':
app.run(debug=True)
FAS Who Plugin for TurboGears2
FASWho Plugin
Authors
Luke Macken Toshio Kuratomi
Date 3 September 2011
This plugin provides authentication to the Fedora Account System using the repoze.who WSGI middleware.
It is designed for use with TurboGears2 but it may be used with any repoze.who using application. Like
jsonfas2, faswho has builtin CSRF protection. This protection is implemented as a second piece of
middleware and may be used with other repoze.who authentication schemes.
Authenticating against FAS with TurboGears2
Setting up authentication against FAS in TurboGears2 is very easy. It requires one change to be made to
app/config/app_cfg.py. This change will take care of registering faswho as the authentication provider,
enabling CSRF protection, switching tg.url() to use fedora.ta2g.utils.url() instead, and allowing the
_csrf_token parameter to be given to any URL.
Using CSRF middleware with other Auth Methods
This section needs to be made clearer so that apps like mirrormanager can be ported to use this.
Templates
The fedora.tg2.utils module contains some templates to help you write CSRF aware login forms and buttons.
You can use the fedora_template() function to integrate them into your templates:
The templates themselves come in two flavors. One set for use with mako and one set for use with genshi.
Mako
Mako version of templates to make adding certain Fedora widgets easier.
fedora.tg2.templates.mako.login.mak
Module author: Toshio Kuratomi <tkuratom@redhat.com>
New in version 0.3.25.
Include this using:
<%namespace name="fedora" file="${context['fedora_template']('login.mak')}" />
fedora.tg2.templates.mako.login.mak.loginform(message='')
kwarg message
Any text or elements contained by the <loginform> tag will be shown as a message to the user.
This is generally used to show status of the last login attempt ("Please provide your
credentials", "Supplied credentials were not correct", etc)
A function for generating the main login form. This is a CSRF token-aware login form that will prompt
for username and password when no session identity is present and ask the user to click a link if they
merely lack a token.
Typical usage, given the above import of the login.mak template would be:
${fedora.loginform()}
fedora.tg2.templates.mako.login.mak.logintoolitem(href=None)
kwarg href
If an href is given, when a user is logged in, their username or display_name will be a link to
the URL.
This function creates an entry into a toolbar for logging into a web app. The entry will contain the
user's username and a logout button if the user is logged in, a verify login button if the user has a
session cookie but not a CSRF token, or a login button if the user doesn't have a session cookie.
Typical usage looks like this:
<ul class="toolbar" id="#main-toolbar">
${fedora.logintoolitem(href=tg.url('/users/info'))}
</ul>
fedora.tg2.templates.mako.jsglobals.mak
Module author: Toshio Kuratomi <tkuratom@redhat.com>
New in version 0.3.25.
Include this using:
<%namespace name="jsglobals" file="${context['fedora_template']('jsglobals.mak')}" />
fedora.tg2.templates.mako.jsglobals.mak.jsglobals()
A function to add global variables to a page. Typically, you'd include this in your master.mak template
and let the javascript variables defined there propagate to every page on your site (since they all
should inherit from master.mak). This adds the following variables in the fedora namespace for other
scripts to access:
fedora.baseurl
URL fragment to prepend to any calls to the application. In a TurboGears application, this is
the scheme, host, and server.webpath. Example: https://admin.fedoraproject.org/pkgdb/. This
may be a relative link.
fedora.identity.anonymous
If true, there will be no other variables in the fedora.identity namespace. If false, these
variables are defined:
fedora.identity.userid
Numeric, unique identifier for the user
fedora.identity.username
Publically visible unique identifier for the user
fedora.identity.display_name
Common human name for the user
fedora.identity.token
csrf token for this user's session to be added to urls that query the server.
Typical usage would be:
${jsglobals.jsglobals()}
Genshi
Genshi version of templates to make adding certain Fedora widgets easier.
fedora.tg2.templates.genshi.login.html
Module author: Toshio Kuratomi <tkuratom@redhat.com>
New in version 0.3.26.
Include this using::
<xi:include href="${fedora_template('login.html', template_type='genshi')}" />
fedora.tg2.templates.genshi.login.html.loginform([message])
message
Any text or elements contained by the <loginform> tag will be shown as a message to the user.
This is generally used to show status of the last login attempt ("Please provide your
credentials", "Supplied credentials were not correct", etc)
A match template for the main login form. This is a CSRF token-aware login form that will prompt for
username and password when no session identity is present and ask the user to click a link if they merely
lack a token.
Typical usage would be:
<loginform>${message}</loginform>
fedora.tg2.templates.genshi.login.html.logintoolitem(@href=URL)
@href If an href attribute is present for this tag, when a user is logged in, their username or
display_name will be a link to the URL.
A match template to add an entry to a toolbar. The entry will contain the user's username and a logout
button if the user is logged in, a verify login button if the user has a session cookie but not a CSRF
token, or a login button if the user doesn't have a session cookie.
Typical usage looks like this:
<ul class="toolbar" id="#main-toolbar">
<logintoolitem href="${tg.url('/users/info')}" />
</ul>
fedora.tg2.templates.genshi.jsglobals.html
Module author: Toshio Kuratomi <tkuratom@redhat.com>
New in version 0.3.26.
Include this using::
<xi:include href="${fedora_template('jsglobals.html', template_type='genshi')}" />
fedora.tg2.templates.genshi.jsglobals.html.jsglobals()
A match template to add global variables to a page. Typically, you'd include this in your master.html
template and let it be added to every other page from there. This adds the following variables in the
fedora namespace for other scripts to access:
fedora.baseurl
URL fragment to prepend to any calls to the application. In a TurboGears application, this is
the scheme, host, and server.webpath. Example: https://admin.fedoraproject.org/pkgdb/. This
may be a relative link.
fedora.identity.anonymous
If true, there will be no other variables in the fedora.identity namespace. If false, these
variables are defined:
fedora.identity.userid
Numeric, unique identifier for the user
fedora.identity.username
Publically visible unique identifier for the user
fedora.identity.display_name
Common human name for the user
fedora.identity.token
csrf token for this user's session to be added to urls that query the server.
Typical usage would be:
<jsglobals />
JAVASCRIPT
Authors
Toshio Kuratomi
Date 26 February 2009
For Version
0.3.x
python-fedora currently provides some JavaScript files to make coding Fedora-Services easier. In the
future we may move these to their own package.
fedora.dojo
Module author: Toshio Kuratomi <tkuratom@redhat.com>
New in version 0.3.10.
Dojo is one of several JavaScript Toolkits. It aims to be a standard library for JavaScript. The
fedora.dojo module has JavaScript code that make use of Dojo to do their work. It is most appropriate to
use when the Dojo libraries are being used as the JavaScript library for the app. However, it is well
namespaced and nothing should prevent it from being used in other apps as well.
API DOCUMENTATION
This API Documentation is currently a catch-all. We're going to merge the API docs into the hand created
docs as we have time to integrate them.
Client
fedora.client is used to interact with Fedora Services.
Changed in version 0.3.21: Deprecate DictContainer in favor of bunch.Bunch
Changed in version 0.3.35: Add the openid clients
Module author: Ricky Zhou <ricky@fedoraproject.org>
Module author: Luke Macken <lmacken@redhat.com>
Module author: Toshio Kuratomi <tkuratom@redhat.com>
exception fedora.client.FedoraServiceError
Base Exception for any problem talking with the Service.
When the Client gets an error talking to the server, an exception of this type is raised. This
can be anything in the networking layer up to an error returned from the server itself.
exception fedora.client.ServerError(url, status, msg)
Unable to talk to the server properly.
This includes network errors and 500 response codes. If the error was generated from an http
response, code is the HTTP response code. Otherwise, code will be -1.
exception fedora.client.AuthError
Error during authentication. For instance, invalid password.
exception fedora.client.AppError(name, message, extras=None)
Error condition that the server is passing back to the client.
exception fedora.client.FedoraClientError
Base Exception for problems which originate within the Clients.
This should be the base class for any exceptions that the Client generates generate. For
instance, if the client performs validation before passing the data on to the Fedora Service.
Problems returned while talking to the Services should be returned via a FedoraServiceError
instead.
exception fedora.client.FASError
FAS Error
exception fedora.client.CLAError
CLA Error
Generic Clients
BaseClient
class fedora.client.BaseClient(base_url, useragent=None, debug=False, insecure=False, username=None,
password=None, httpauth=None, session_cookie=None, session_id=None, session_name='tg-visit',
cache_session=True, retries=None, timeout=None)
A client for interacting with web services.
logout()
Logout from the server.
send_request(method, req_params=None, file_params=None, auth=False, retries=None, timeout=None,
**kwargs)
Make an HTTP request to a server method.
The given method is called with any parameters set in req_params. If auth is True, then
the request is made with an authenticated session cookie.
Parameters
• method -- Method to call on the server. It's a url fragment that comes after the
base_url set in __init__().
• req_params -- Extra parameters to send to the server.
• file_params -- dict of files where the key is the name of the file field used in
the remote method and the value is the local path of the file to be uploaded. If
you want to pass multiple files to a single file field, pass the paths as a list
of paths.
• auth -- If True perform auth to the server, else do not.
• retries -- if we get an unknown or possibly transient error from the server, retry
this many times. Setting this to a negative number makes it try forever. Default
to use the retries value set on the instance or in __init__() (which defaults to
zero, no retries).
• timeout -- A float describing the timeout of the connection. The timeout only
affects the connection process itself, not the downloading of the response body.
Default to use the timeout value set on the instance or in __init__() (which
defaults to 120s).
Return type
Bunch
Returns
The data from the server
Changed in version 0.3.21: * Return data as a Bunch instead of a DictContainer * Add
file_params to allow uploading files
Changed in version 0.3.33: * Added the timeout kwarg
session_cookie
Deprecated, use session_id instead.
The session cookie is saved in a file in case it is needed in consecutive runs of
BaseClient.
session_id
The session_id.
The session id is saved in a file in case it is needed in consecutive runs of BaseClient.
ProxyClient
class fedora.client.ProxyClient(base_url, useragent=None, session_name='tg-visit',
session_as_cookie=True, debug=False, insecure=False, retries=None, timeout=None)
A client to a Fedora Service. This class is optimized to proxy multiple users to a service.
ProxyClient is designed to be threadsafe so that code can instantiate one instance of the class
and use it for multiple requests for different users from different threads.
If you want something that can manage one user's connection to a Fedora Service, then look into
using BaseClient instead.
This class has several attributes. These may be changed after instantiation however, please note
that this class is intended to be threadsafe. Changing these values when another thread may
affect more than just the thread that you are making the change in. (For instance, changing the
debug option could cause other threads to start logging debug messages in the middle of a method.)
base_url
Initial portion of the url to contact the server. It is highly recommended not to change
this value unless you know that no other threads are accessing this ProxyClient instance.
useragent
Changes the useragent string that is reported to the web server.
session_name
Name of the cookie that holds the authentication value.
session_as_cookie
If True, then the session information is saved locally as a cookie. This is here for
backwards compatibility. New code should set this to False when constructing the
ProxyClient.
debug If True, then more verbose logging is performed to aid in debugging issues.
insecure
If True then the connection to the server is not checked to be sure that any SSL
certificate information is valid. That means that a remote host can lie about who it is.
Useful for development but should not be used in production code.
retries
Setting this to a positive integer will retry failed requests to the web server this many
times. Setting to a negative integer will retry forever.
timeout
A float describing the timeout of the connection. The timeout only
affects the connection process itself, not the downloading of the
response body. Defaults to 120 seconds.
Changed in version 0.3.33: Added the timeout attribute
debug When True, we log extra debugging statements. When False, we only log errors.
log = <logging.Logger object>
send_request(method, req_params=None, auth_params=None, file_params=None, retries=None,
timeout=None)
Make an HTTP request to a server method.
The given method is called with any parameters set in req_params. If auth is True, then
the request is made with an authenticated session cookie. Note that path parameters should
be set by adding onto the method, not via req_params.
Parameters
• method -- Method to call on the server. It's a url fragment that comes after the
base_url set in __init__(). Note that any parameters set as extra path
information should be listed here, not in req_params.
• req_params -- dict containing extra parameters to send to the server
• auth_params --
dict containing one or more means of authenticating to the server. Valid entries
in this dict are:
cookie Deprecated Use session_id instead. If both cookie and session_id are set,
only session_id will be used. A Cookie.SimpleCookie to send as a session
cookie to the server
session_id
Session id to put in a cookie to construct an identity for the server
username
Username to send to the server
password
Password to use with username to send to the server
httpauth
If set to basic then use HTTP Basic Authentication to send the username and
password to the server. This may be extended in the future to support
other httpauth types than basic.
Note that cookie can be sent alone but if one of username or password is set the
other must as well. Code can set all of these if it wants and all of them will be
sent to the server. Be careful of sending cookies that do not match with the
username in this case as the server can decide what to do in this case.
• file_params -- dict of files where the key is the name of the file field used in
the remote method and the value is the local path of the file to be uploaded. If
you want to pass multiple files to a single file field, pass the paths as a list
of paths.
• retries -- if we get an unknown or possibly transient error from the server, retry
this many times. Setting this to a negative number makes it try forever. Default
to use the retries value set on the instance or in __init__().
• timeout -- A float describing the timeout of the connection. The timeout only
affects the connection process itself, not the downloading of the response body.
Defaults to the timeout value set on the instance or in __init__().
Returns
If ProxyClient is created with session_as_cookie=True (the default), a tuple of
session cookie and data from the server. If ProxyClient was created with
session_as_cookie=False, a tuple of session_id and data instead.
Return type
tuple of session information and data from server
Changed in version 0.3.17: No longer send tg_format=json parameter. We rely solely on the
Accept: application/json header now.
Changed in version 0.3.21: * Return data as a Bunch instead of a DictContainer * Add
file_params to allow uploading files
Changed in version 0.3.33: Added the timeout kwarg
OpenIdBaseClient
class fedora.client.OpenIdBaseClient(base_url, login_url=None, useragent=None, debug=False,
insecure=False, openid_insecure=False, username=None, cache_session=True, retries=None, timeout=None,
retry_backoff_factor=0)
A client for interacting with web services relying on openid auth.
has_cookies()
login(username, password, otp=None)
Open a session for the user.
Log in the user with the specified username and password against the FAS OpenID server.
Parameters
• username -- the FAS username of the user that wants to log in
• password -- the FAS password of the user that wants to log in
• otp -- currently unused. Eventually a way to send an otp to the API that the API
can use.
send_request(method, auth=False, verb='POST', **kwargs)
Make an HTTP request to a server method.
The given method is called with any parameters set in req_params. If auth is True, then
the request is made with an authenticated session cookie.
Parameters
• method -- Method to call on the server. It's a url fragment that comes after the
base_url set in __init__().
• auth -- If True perform auth to the server, else do not.
• req_params -- Extra parameters to send to the server.
• file_params -- dict of files where the key is the name of the file field used in
the remote method and the value is the local path of the file to be uploaded. If
you want to pass multiple files to a single file field, pass the paths as a list
of paths.
• verb -- HTTP verb to use. GET and POST are currently supported. POST is the
default.
session_key
fedora.client.openidbaseclient.requires_login(func)
Decorator function for get or post requests requiring login.
Decorate a controller method that requires the user to be authenticated. Example:
from fedora.client.openidbaseclient import requires_login
@requires_login
def rename_user(new_name):
user = new_name
# [...]
OpenIdProxyClient
class fedora.client.OpenIdProxyClient(base_url, login_url=None, useragent=None, session_name='session',
debug=False, insecure=False, openid_insecure=False, retries=None, timeout=None)
A client to a Fedora Service. This class is optimized to proxy multiple users to a service.
OpenIdProxyClient is designed to be usable by code that creates a single instance of this class
and uses it in multiple threads. However it is not completely threadsafe. See the information on
setting attributes below.
If you want something that can manage one user's connection to a Fedora Service, then look into
using OpenIdBaseClient instead.
This class has several attributes. These may be changed after instantiation. Please note,
however, that changing these values when another thread is utilizing the same instance may affect
more than just the thread that you are making the change in. (For instance, changing the debug
option could cause other threads to start logging debug messages in the middle of a method.)
base_url
Initial portion of the url to contact the server. It is highly recommended not to change
this value unless you know that no other threads are accessing this OpenIdProxyClient
instance.
useragent
Changes the useragent string that is reported to the web server.
session_name
Name of the cookie that holds the authentication value.
debug If True, then more verbose logging is performed to aid in debugging issues.
insecure
If True then the connection to the server is not checked to be sure that any SSL
certificate information is valid. That means that a remote host can lie about who it is.
Useful for development but should not be used in production code.
retries
Setting this to a positive integer will retry failed requests to the web server this many
times. Setting to a negative integer will retry forever.
timeout
A float describing the timeout of the connection. The timeout only affects the connection
process itself, not the downloading of the response body. Defaults to 120 seconds.
debug When True, we log extra debugging statements. When False, we only log errors.
login(username, password, otp=None)
Open a session for the user.
Log in the user with the specified username and password against the FAS OpenID server.
Parameters
• username -- the FAS username of the user that wants to log in
• password -- the FAS password of the user that wants to log in
• otp -- currently unused. Eventually a way to send an otp to the API that the API
can use.
Returns
a tuple containing both the response from the OpenID provider and the session used
to by this provider.
send_request(method, verb='POST', req_params=None, auth_params=None, file_params=None,
retries=None, timeout=None, headers=None)
Make an HTTP request to a server method.
The given method is called with any parameters set in req_params. If auth is True, then
the request is made with an authenticated session cookie. Note that path parameters should
be set by adding onto the method, not via req_params.
Parameters
• method -- Method to call on the server. It's a url fragment that comes after the
base_url set in __init__(). Note that any parameters set as extra path
information should be listed here, not in req_params.
• req_params -- dict containing extra parameters to send to the server
• auth_params --
dict containing one or more means of authenticating to the server. Valid entries
in this dict are:
cookie Deprecated Use session_id instead. If both cookie and session_id are set,
only session_id will be used. A Cookie.SimpleCookie to send as a session
cookie to the server
session_id
Session id to put in a cookie to construct an identity for the server
username
Username to send to the server
password
Password to use with username to send to the server
httpauth
If set to basic then use HTTP Basic Authentication to send the username and
password to the server. This may be extended in the future to support
other httpauth types than basic.
Note that cookie can be sent alone but if one of username or password is set the
other must as well. Code can set all of these if it wants and all of them will be
sent to the server. Be careful of sending cookies that do not match with the
username in this case as the server can decide what to do in this case.
• file_params -- dict of files where the key is the name of the file field used in
the remote method and the value is the local path of the file to be uploaded. If
you want to pass multiple files to a single file field, pass the paths as a list
of paths.
• retries -- if we get an unknown or possibly transient error from the server, retry
this many times. Setting this to a negative number makes it try forever. Default
to use the retries value set on the instance or in __init__().
• timeout -- A float describing the timeout of the connection. The timeout only
affects the connection process itself, not the downloading of the response body.
Defaults to the timeout value set on the instance or in __init__().
• headers -- A dictionary containing specific headers to add to the request made.
Returns
A tuple of session_id and data.
Return type
tuple of session information and data from server
Clients for Specific Services
Wiki
class fedora.client.Wiki(base_url='https://fedoraproject.org/w/', *args, **kwargs)
api_high_limits = False
check_api_limits()
Checks whether you have the 'apihighlimits' right or not.
fetch_all_revisions(start=1, flags=True, timestamp=True, user=True, size=False, comment=True,
content=False, title=True, ignore_imported_revs=True, ignore_wikibot=False, callback=None)
Fetch data for all revisions. This could take a long time. You can start at a specific
revision by modifying the 'start' keyword argument.
To ignore revisions made by "ImportUser" and "Admin" set ignore_imported_revs to True (this
is the default). To ignore edits made by Wikibot set ignore_wikibot to True (False is the
default).
Modifying the remainder of the keyword arguments will return less/more data.
get_recent_changes(now, then, limit=500)
Get recent wiki changes from now until then
login(username, password)
print_recent_changes(days=7, show=10)
Service
Transforming SQLAlchemy Objects into JSON
GLOSSARY
controller
In MVC design, the controller is in charge of things. It takes processes events and decides what
data to ask the model for, manipulates the data according to the information in the event, and
decides which view to send the results to to be rendered.
CSRF Cross-site request forgery is a technique where a malicious website can gain access to another web
site by hijaacking a currently open session that the user has open to the site. This technique
can also affect identification via SSL Certificates or anything else that the browser sends to the
server automatically when a request is made.
SEE ALSO:
CSRF-Protection
Dojo Dojo is a JavaScript toolkit that aims to be a standard library for JavaScript. It provides a
small core library with useful functions and an expanded set of scripts that can be added that
provide widgets and other features.
SEE ALSO:
http://www.dojotoolkit.org
double submit
A strategy to foil CSRF attacks. This strategy involves sending the value of the authentication
cookie (or something derivable only from knowing the value of the authentication cookie) in the
body of the request. Since the Same Origin Policy prevents a web site other than the one
originating the cookie from reading what's in the cookie, the server can be reasonably assured
that the request does not originate from an unknown request on another website. Note that this
and other anti-CSRF measures do not protect against spoofing or getting a user to actively click
on a link on an attacked website by mistake.
flask A simple Python web framework that we're using in parts of Fedora Infrastructure. It provides
good documentation and simplicity in its design.
SEE ALSO:
http://flask.pocoo.org/docs/
JSON JavaScript Object Notation is a format for marshalling data. It is based on a subset of
JavaScript that is used to declare objects. Compared to xml, JSON is a lightweight, easily parsed
format.
SEE ALSO:
Wikipedia's JSON Entry
model In MVC design, the layer that deals directly with the data.
OpenID A specification for single sign on to web services where the authentication server and the
application seeking to have the user authenticated do not need to have complete trust in each
other.
SEE ALSO:
http://openid.net/get-an-openid/what-is-openid/
Same Origin Policy
A web browser security policy that prevents one website from reading: 1) the cookies from another
website 2) the response body from another website
SEE ALSO:
http://en.wikipedia.org/wiki/Same_origin_policy
single sign-on
A feature that allows one login to authenticate a user for multiple applications. So logging into
one application will authenticate you for all the applications that support the same
single-sign-on infrastructure.
TurboGears
A Python web framework that most of Fedora Infrastructure's apps are built on.
SEE ALSO:
http://www.turbogears.org/
TurboGears2
The successor to TurboGears, TurboGears2 provides a very similar framework to coders but has some
notable differences. It is based on pylons and paste so it is much more tightly integrated with
WSGI. The differences with :ref`TurboGears`1 are largely with the organization of code and how to
configure the application.
SEE ALSO:
http://www.turbogears.org/
view In MVC design, the layer that takes care of formatting and rendering data for the consumer. This
could be displaying the data as an html page or marshalling it into JSON objects.
WSGI WSGI is an interface between web servers and web frameworks that originated in the Python
community. WSGI lets different components embed each other even if they were originally written
for different python web frameworks.
SEE ALSO:
http://en.wikipedia.org/wiki/Web_Server_Gateway_Interface
• glossary
• genindex
• modindex
• search
COPYRIGHT
2007-2018 Red Hat, Inc.