It is a complicated almost religious discussion. I don't think assertions (SAML, openID) or PKI client Auth are going away any time soon. The trust model and technology are slightly disjoint discussions. People can use smartcards with self issued certificates and SAML providers can sign assertions with certificates with roots ross certified to specific PKI bridges. This is probably more a FIWG discussion. That is where the rubber hits the road for the more complicated cross federation issues, when it comes to trust models. For certificates vs assertions there is a privacy related issue. Depending on the use case. For Defence intelligence and Police credentials there may be no expectation of privacy or anonymity or privacy when your credential is used. However many privacy people in the Citizen to Gov use case want to stop correlating identifiers across sites. In some case there is a legal requirement for this. I helped start Xcert software (now RSA KeyOn) 12 years ago to work on federated identity issues using PKI client Auth. Why PKI failed in the consumer/internet space is a big topic. In the US FICAM anticipates the vast majority of external credentials it will accept to be assertion based. I should also mention that u-prove (zero knowledge prrof) cryptography contains elements of both certificates and assertions. I have limited expectations for any short term traction on that however. The reality is that the main driving force on the internet is access to API, and attributes. SSO is just something that is going along for the ride. The US Gov PIV card deployment needs BAE (SAML attribute query) to retrieve attributes and be useful. (perhaps overstated) Lets discuss how you want to separate the issues. If we tackle them all together we will probably get nowhere. John B. On 2011-03-14, at 8:08 AM, Rainer Hörbe wrote:
John, Patrick and I had a discussion about the pros and cons of federation models based on credentials versus assertions. The attached document is a preliminary result with conclusions like PKI and non-PKI federation models need to be combined in most cases at higher LoA To implement a federation an RFC 3647-style policy is insufficient; A more complete Trust Framework is needed Whereas the Higher Education sector favors brokered trust, e-Government and Industry prefer the PKI approach. But it is not a question of one way or the other.
Request for feedback: I wonder where this discussion should be homed. FIWG, BCTF and TFMM are related, and it is also an extrakantarian issue. Any interest to take over this discussion?
- Rainer <pki vs non-pki.pdf>