Let's take this from the top. XAuth is a proposal to let browsers remember that sites have registered themselves as a user's identity provider and let other sites know if the user has a session at that site. In other words, it has the same information as proprietary solutions that already exist, except that it works across multiple identity providers. It means that when you go to a new website, it doesn't have to ask you what your preferred services are, it can just look them up. Note that this only tells the site that you have an account with Google or Yahoo or Facebook or Twitter, not what the account is or any private information.
Some objections to all of this are natural; I had them too. Below are my answers (note: I'm not speaking for Google here, just myself, but I believe these answers are consistent with what people like Chris Messina and Joseph Smarr are thinking).
- Objection: The implementation relies on a single domain. Answer: The current implementation does this due to restrictions in browser security models. There is no essential reason for that single domain to exist, there is no data stored on the domain's servers and no web services accessed from it. It just stores the JS used to implement the API today. In theory one could write a browser extension that maps "xauth.org" to "127.0.0.1", make sure to serve up the JS from a local web server, and XAuth would work fine. A better use of time is to convince the browser vendors to support XAuth natively. Which leads to objection #2...
- Objection: This should be done by the browsers, not by a back door via a JS API. Answer: Yes, it absolutely should, and the best and fastest way to make that happen is to get a deployed solution and user experience supported by many sites as quickly as possible, then swap out the implementation. It is trivial to replace the XAuth JS core with calls to a browser solution. In fact it'd be great to figure out a good way to make this happen automatically with feature sniffing, though for that we'd need a very stable JS interface. And for that we need... field deployment experience! XAuth is a great bootstrap to a browser centric future solution.
- Objection: But the central domain name should not be managed by a single party, even in the interim: Answer: Yes, that's true, and the parties involved are working to get it managed by a neutral third party who can be trusted not to do something bad. Note that we can know fairly easily if some domain owner starts to do bad things as they'd have to mess around with the publicly visible JS to get information to go anywhere but the users' local computer.
- Objection: It should be opt-in per browser, not opt-out. Answer: Note that XAuth doesn't prevent opt-in semantics on the part of identity providers, it just doesn't enforce it. All other things being equal, yes, it would be better to force the default to "off" and ensure users understand what they're doing when they turn it "on". The last bit is hard; a choice that users don't understand is even worse; it doesn't really solve anything but hurts adoption -- and gives proprietary alternatives, which don't require opt-in, a big advantage. Again, a browser-native alternative may be able to solve this. An XAuth ecosystem, already in place, ready for browsers to use, is a powerful incentive for them to do this.
Eran writes: "This is a misguided attempt to solve a problem browser vendors have failed to address. It is true that getting browser vendors to care about identity and innovate in the space is a huge challenge, but solving it with a server-hosted, centralized solution goes against everything the distributed identity movement has tried to accomplish over the past few years." (Emphasis mine.) It is precisely my hope that XAuth will help bootstrap the ultimate goals of the distributed identity movement, and that it will actually accomplish things in a short time frame while setting the environment up for a better solution in the future.
[Updated: You may want to refer to the original announcement of XAuth back in April, where many of these objections have already been answered.]
[Updated: You may want to refer to the original announcement of XAuth back in April, where many of these objections have already been answered.]
Instead of XAuth, why not just use a two stage security platform, like a mobile phone, to provide OTPs. The user could use the same user name and password, plus a randomly generated OTP for each login and be done with it. Problem with OpenID, and XAuth by proxy, is that theft of a single ID could compromise every service attached to that ID.
ReplyDeleteSince the majority of users re-use their username (email) and password across dozens of sites today, theft of a single ID already compromises every service the user uses. It's just much, much harder to recover from.
ReplyDeleteXAuth doesn't do authentication, it just points to your authentication provider. You can then choose to use an authentication provider that does OTP.
It's not something browsers "failed to address". We've had client certs since Netscape 3 and they work fine. It's an identity, you create them youself, and you can get encrypted communications for free.
ReplyDeleteGranted, it could be easier to use. But any browser could improve their UI in that area without affecting any services already in use.
Just my .02 (and really criticism against openid, not xauth)
It is strange to call items 2-4 "answers" when
ReplyDeleteyou seem to be agreeing with the objections.
In light of this it may be unfair to call Eran's
post on XAuth "overblown".
Very interesting topic.. However the idea of storing the users' identity provider in the browser seems a bit odd. That would mean all browsers (and older versions) would have to provide some facility to allow for such user provider registration in which case such information will most likely be stored as a session cookie (or perhaps in some other form dictated by a new RFC that all browsers have to support (highly unlikely)). If this is supported via cookies then what if cookies are disabled on the browser? Would the user be prompted to enter the authentication details of every single domain when authentication fails? The other approach that comes to mind would be to use a common trusted identity and access management (IAM) service provider that binds together the user session information across different service providers pretty much like a SSO except it is across different service provider domains -- Arun, P
ReplyDelete