- From: Thomas Roessler <tlr@w3.org>
- Date: Mon, 19 Jun 2006 21:06:00 +0200
- To: public-ietf-w3c@w3.org
As a follow-up to the question that came up during today's call. Regards, -- Thomas Roessler, W3C <tlr@w3.org> ----- Forwarded message from Sam Hartman <hartmans-ietf@mit.edu> ----- From: Sam Hartman <hartmans-ietf@mit.edu> To: dix@ietf.org Date: Mon, 19 Jun 2006 13:31:21 -0400 Subject: [dix] An HTTP-based solution for digital identity Reply-To: Digital Identity Exchange <dix@ietf.org> List-Id: Digital Identity Exchange <dix.ietf.org> X-Spam-Level: Hi, folks. I said I was working on a draft describing use cases and a specific technical proposal. I have that draft done and it compliments my phishing requirements draft. I expect to update the phishing requirements draft based on several reviews I've received and based on what confuses people about it when I have explained it to them in person. I was intrigued by Phil Hallam-Baker's idea to separate HTTP authentication into two parts : one between a user and some authentication/identity service and one between the authentication service and the relying party. It is important to standardize the second protocol; while standard methods of authentication to a identity service would be desirable extensions offered by specific identity providers may be useful and can be supported. Phil suggested starting with something like http-digest. I started doing that mentally during the last IETF meeting. I found that I was reinventing a lot of the Kerberos ap-req exchange--the exchange used to authenticate to the server. I started thinking about reuse of deployed code and realized that it would probably be desirable to use existing code rather than inventing some new protocol. New protocols are often more complicated than they at first seem they need to be and I definitely think this would be on exception. So, I decided to put together a sketch of how you could just use Kerberos to meet these needs. I was shocked at how much of the infrastructure necessary to do this is already available in current web browsers. So, I decided to write up a proposal on how to realize some of the goals in this space using Kerberos and other related technologies. I also described the use cases I'm attempting to solve. My use cases are less ambitious than the DIX use cases; personally I think that is a good thing. I do believe that my proposal can be expanded to meet all the use cases that impinge on protocol interactions. My belief is that this should be done gradually. I'm sure someone out there is going to claim that Kerberos is a horrible fit for this and that it is too complicated. I encourage you to specify an alternative in at least as much detail as I have done so we can compare the complexity, functionality, reuse of technology and ease of deployment of proposals. While any comments are welcome, alternative proposals or explanations of how I got use cases wrong may make constructive discussion easier than comments of the form "this sucks." At some level, the DIX proposal is an alternative. However as I think more about this problem space, I'm coming to the conclusion that a redirect based approach is not really in conflict with a longer-term in-band approach. My approach will work better for webdav, caldav and other non-HTML-based approaches. I think my approach may provide a better user experience even for web pages once it is available. However the DIX proposal can be deployed with much less effort. If both proposals end up using SAML to express information about identity, they could be complimentary. I do have some serious questions about whether a redirect-based approach should be standardized. I'll ask those questions in another message. I'm also talking to people interested in deploying redirect-based approaches along with those who have done so, trying to get a more informed opinion. Any way, I'd like to draw your attention to draft-hartman-webauth-00.txt. Let the flames begin. _______________________________________________ dix mailing list dix@ietf.org https://www1.ietf.org/mailman/listinfo/dix ----- End forwarded message -----
Received on Monday, 19 June 2006 19:06:06 UTC