All the tests that require no more than the standard library pass -- however modules that depend on external packages (such as the pyCURL fetcher or the MySQL / PostgreSQL stores) haven't been updated and may not work.
This implementation of OpenID has been ported to Python 3 -- all but one test is known to pass.
This implementation of OpenID has been upgraded to support version 2.0 of the OpenID Authentication specification.
New in this version is:
-
Verification of relying party return_to addresses, to screen out RPs hiding behind open redirect relays. Server code can invoke this with the returnToVerified method on CheckIDRequest.
-
Helper module for the Provider Authentication Policy Extension (PAPE) in openid.extensions.pape.
-
Helper module for Attribute Exchange in openid.extensions.ax.
Bugfixes:
-
Allow the use of lxml as an ElementTree implemenation.
-
Provide compatability with a wider range of versions for SQL stores.
The third argument to Consumer.complete() is required.
The sreg module should be imported from openid.extensions.sreg instead of openid.sreg.
The ax module should likewise be imported from openid.extensions.ax instead of openid.ax
The openid.extensions.ax.FetchRequest.fromOpenIDRequest method now takes a CheckIDRequest object instead of a Message object
The OpenID response (the result of Consumer.complete()) now has a getDisplayIdentifier() method which should be called instead of accessing response.identity_url. The value of getDisplayIdentifier() will be the XRI i-name if XRI is used. The value of response.identity_url SHOULD, however, be used as the application's database key for storing account information.
The big news here is compatibility with svn revision 313 of the OpenID 2.0 draft specification.
Highlights include:
- Simple Registration support in a new module openid.sreg. (Those previously using SuccessResponse.extensionResponse are advised to look here.)
- OpenID provider-driven identifier selection.
- "Negotiators" allow you to define which association types to use.
- Examples for Django.
Python 2.5 is now supported. Support for Python 2.2 discontinued. Seperate installation of yadis and urljr packages is no longer required; they have been included in this package.
One of the additions to the OpenID protocol was a specified nonce format for one-way nonces. As a result, the nonce table in the store has changed. You'll need to run contrib/upgrade-store-1.1-to-2.0 to upgrade your store, or you'll encounter errors about the wrong number of columns in the oid_nonces table.
If you've written your own custom store or code that interacts directly with it, you'll need to review the change notes in openid.store.interface.
Consumers should now pass an additional parameter to Consumer.complete() to defend against return_to tampering.
This version of the library allows the use of XRI as OpenID identifiers, allowing users to log in with their i-names. For full XRI compatibility, relying parties integrating this library should take note of the user's CanonicalID, as described in the "Identifying the End User" section of the OpenID 2.0 specification.
A variety of bug fixes were included in this release, mostly relating to international issues such as dealing with other character sets, Unicode, incorrectly flagging certain Norwegian trust roots as suspect, and operation of the filesystem-backed store on exotic platforms.
- urljr 1.0.1
- yadis 1.1.0
Version 1.1 of the Python OpenID library implements recent changes to the OpenID specification as well as making API changes that should make integration with applications easier.
One of the major changes to OpenID since the last release has been the approval of Yadis discovery as the preferred way to specify the OpenID metadata for an identity URL instead of using tags in HTML. This library does Yadis discovery, and if that fails, it falls back to old-style discovery.
Some advantages of Yadis support are:
-
Support for fallback if your primary OpenID provider is not available
-
Support for load-balancing between OpenID servers
-
Easy interoperability for different identity services
For more information about Yadis, see http://yadis.org/
OpenID also has formalized support for extensions. Extensions are a mechanism for transferring information from the consumer to the server and from the server to the consumer in the process of performing OpenID authentication. Extensions are implemented as additional namespaced query arguments that go along with standard OpenID requests and responses. This library provides a simple API for adding extension arguments to requests and extracting extension responses from replies.
These dependencies should be available from wherever you acquired the OpenID library.
-
urljr - The fetcher abstraction from the previous OpenID library has been extended and is also used for the Yadis library. Because the Yadis library is useful without the OpenID library, the HTTP fetching code has been rolled into its own package. Additionally, the library now has the concept of a default fetcher to make APIs simpler.
-
yadis - The Yadis library provides a general discovery layer that has been adopted by OpenID as well as other identity-related protocols. Most OpenID identity URLs will work without the Yadis library, but as time goes on, this library will be more and more important.
The consumer API has been changed for more natural use as well as to support extension arguments.
- OpenIDConsumer(store, [fetcher], [immediate]) is now Consumer(session, store)
-
The session object is a dictionary-like object that should be tied to the requesting HTTP agent, for example, using a session ID cookie. It is used for Yadis fallback and holding the state of the OpenID transaction between the redirect to the server and the response. The values that are placed in the session are namespaced, so there should not be a conflict with other uses of the same session. The session namespace is an attribute of the Consumer object.
-
Because the consumer object now does session management, it is necessary to construct a new consumer object for every request. Creating consumer objects is light-weight.
-
OpenIDConsumer.beginAuth(user_url) is now Consumer.begin(user_url) and either returns an AuthRequest object or raises an exception. There is no more tuple unpacking or status codes.
-
OpenIDConsumer.constructRedirect(authreq, return_to, trust_root) is now AuthRequest.redirectURL(trust_root, return_to, [immediate]).
-
OpenIDConsumer.completeAuth(token, query) is now Consumer.complete(query). It no longer returns a tuple. Instead it returns an object that has a status code and additional information about the response. See the API documentation for more information.
The server API has been changed for greater extensibility. Instead of taking an "is_authorized" callback, processing happens in several stages, allowing you to insert extension data into the response before it is signed and returned. See the documentation for the openid.server.server module.
- fetcher was openid.consumer.fetchers.OpenIDHTTPFetcher, is now urljr.fetchers.HTTPFetcher. get() and post() have been replaced by fetch(), see urljr.fetchers for details.
The server changed the way it indexes associations in the store, so if you're upgrading a server installation, we recommend you clear the old records from your store when you do so. As a consequence, consumers will re-establish associations with your server a little sooner than they would have otherwise.