Most people have probably seen or used OpenID. If you have used it, then it has most likely that it was with the 1.x protocol. Now that OpenID 2.0 is close to release (apparently they really mean it this time …), it is worth looking at the new features it enables. A few that have stood out to me include:
- proper extension support
- support for larger requests/responses
- directed identity
- attribute exchange extension
- support for a new naming monopoly
I’ll now discuss each of these in a bit more detail
OpenID 1.1 had one well known extension: the Simple Registration Extension. An OpenID relying party (RP) would send a request with an openid.sreg.required field, and get back user information in openid.sreg.* fields from the OpenID Provider (OP). The RP and OP would just need to know that “openid.sreg” fields means that the simple registration extension is being used.
But what if I want to define my own extension? If my RP sends openid.foo.* fields, how does the OP know that it refers to my extension and not some other extension that happened to pick the same prefix?
OpenID 2.0 solves this problem by borrowing the idea of name space URIs from XML. If I am sending some openid.foo.* fields in an OpenID message, then I also need to send an openid.ns.foo field set to a URI that identifies the extension. This means that a message that sends the same data as openid.bar.* fields should be treated the same provided that openid.ns.bar is set to the extension’s name space URI.
As with XML name spaces, this allows us to piggy back on top of DNS as a way of avoiding conflicts.
Large Requests and Responses
OpenID 1.1 uses HTTP redirects as a way of transferring control between the RP and OP (and vice versa). This means that the upper limit on a message is effectively the same as the smallest upper limit on length of URLs in common web browsers and servers. Internet Explorer seems to have the lowest limit—2,083 characters—so it sets the effective limit on message size.
For simple authentication checks (what OpenID was originally designed for), this is not generally a problem. But once you start to introduce a few extensions, this limit can easily be reached.
For OpenID 1.1, the authentication process goes something like this:
- the user enters their identity URL into a form on the RP
- the RP performs discovery on that URL to find the user’s OP.
- the RP initiates an OpenID authentication request with that OP.
With OpenID 2.0, the discovery process may tell the RP that the URL identifies the OP rather than the user. If this happens, the RP proceeds with the authentication request using the special “http://specs.openid.net/auth/2.0/identifier_select” value as the identity URL. The OP will then fill in the user’s actual identity URL in the subsequent authentication response. As an additional step, the RP is then required to perform discovery on this URL to ensure that the OP is entitled to authenticate it.
There are a number of cases where this feature can be useful:
- An OpenID provider can give their users a single URL that will work for everyone. For instance, if AOL sets things up correctly, you’d be able to type “aol.com” into any OpenID 2.0 enabled site to log in with an AIM screen name.
- A privacy concious user could configure their own OpenID provider that will hand out different identity URLs to different RPs, similar to how some people use single-purpose email addresses today.
- If an RP requires that users use a particular OP, they could use directed identity to begin the authentication request without requiring the user to enter an identity URL.
Attribute Exchange Extension
The OpenID Attribute Exchange extension is like the simple registration extension on steroids. The major differences are:
- Unlike the simple registration extension, the attribute exchange extension does not have a fixed set of attributes that can be transmitted. Instead it uses URIs to identify the attribute types, making it easy to define new attributes without causing conflicts. Of course an attribute is not particularly useful if no one else supports it, so there is a process set up to standardise common attribute types.
- As well as receiving attribute values as part of an authentication response, an RP can request that an OP store certain attribute values. This is done as part of an authentication request, so the OP can verify that the user really wants to store the values.
- The RP can request ongoing updates for the attributes it is interested in. As an example, if you stored your hackergotchi with your OP, changes to the image could be automatically pushed out to all sites you use that want to display that image.
Prop Up A New Naming Monopoly
With OpenID 2.0, a user is supposed to be able to enter an i-name in place of an identity URL in an RP, and be authenticated against the i-broker managing that name. So rather than entering an ugly URL, users can enter an ugly string starting with “=” or “@”.
All it costs to take advantage of this is US$12 per year (or US$55 for an organisation name). They claim that it will be possible to use an i-name in many contexts in the future, but for now it appears to be limited to (1) a subset of OpenID RPs, (2) a web form that people can use to send you emails and (3) an HTTP redirection to your website.
At this point, it seems that i-name support in OpenID is more important to the i-name crowd than the OpenID crowd. That said, the complexity is hidden by most of the existing OpenID libraries, so it’ll most likely get implemented by default on most RPs moving forward.
Overall OpenID 2.0 looks like a worthwhile upgrade, even if some parts like i-names are questionable.
Assuming the attribute exchange extension takes off, it should provide a much richer user experience. Imagine being able to update your shipping address in one place when you move house and having all the online retailers you use receive the updated address immediately. Or changing your email address and having all the bugzilla instances you use pick up the new address instantly (perhaps requiring you to verify the new address first, of course).
The improved extension support should also make it easier for people to experiment with new extensions without accidentally conflicting with each other, which should accelerate development of new features.