Curtis, Sorry for the long response. I appreciate that you were specifically looking for an IETF response to you posting - I wanted to give you my CyberSafe response for what it is worth. I hope you can give me feedback on my opinion/comments.
There is a general market missunderstanding/confusion between PKI and Kerberos authentication technology - I beleive this is mostly caused by the large amount of marketing around PKI and related products/companies over the last few years - many companies thought that PKI was the only authentication solution they would ever need. In my view what has happened is that the adoption of Kerberos by Microsoft in Win2k/XP and .NET has allowed many organisations to now recognise that Kerberos can be deployed in the enterprise network and it is being considered more seriously than before as a viable technology for their network security and authentication needs. I have been able to verify this view with the clear interest we have seen over the last 12-18 months in our current products and services and also many from feedback received from customers/prospects/partners when asked for their views. It is also worth highlighting that Kerberos is often positioned as an authentication technology, which of course is correct, but it is also becoming widely chosen for its symmetric key management capabilities. If you combine this key management with PK authentication to give two-factor authenticated access to symmetric keys with the private key from the PK being used to give a digital signature capability then the opportunity is very exciting. I am sure you are also aware that Kerberos can be extended using pre-authentication to use other two-factor authentication tokens and methods. I know of two global solutions where the approach of using key management and PK together is strategically utilised - one in the Cable Network industry (www.packetcable.com) and another is in the Wireless LAN (WLAN) industry where Kerberos is becoming utilised for key management, but a client certificate is still used so that a user at their workstation can sign documents and use two-factor authentication (stronger than regular userid/password). It is also worth mentioning that Kerberos has strengths when compared to alternatives in implementation of a standards based application security solution - I can explain this to you more if you are interested. It means that an increasing number of vendors are exploring use of Kerberos credentials to secure their application and allow improved authentication/integrity and confidentiality. Anyway, I would like to get back to your specific questions and try and help you with your confusion. I hope I don't confuse you further ? It is our opinion that the future of Kerberos is going to rely more on PK than today - the best of each technology is gradualy being realised and 'merged'. I know this view is also shared by Microsoft. The PKINIT extension to Kerberos has been available for many years as a draft and is nearly reaching RFC status with IETF. We are seeing an increasing number of examples where companies are realising the strengths of Kerberos and strengths of PK and utilising them together with PKINIT to get the best of both worlds. We also need to consider PKCROSS and other PK related extensions in this discussion and recognise that other PK related extensions to Kerberos may become widely used in the future. From a pure technology point of view it is clear that the PK advantage (depending on what you are trying to acheive) is that PK needs a private key - this private key can be used for digital signatures. There are also disadvantages of PK technology, but there are with many technologies. There are many comparisons that can be drawn between PK and Kerberos infrastructures and it quickly becomes apparent that they are actually very complementary. Anyway, if you are looking for a single common authentication solution then you need to recognise that Kerberos is allready a standard for initial operating system login for many operating system vendors, both Microsoft and UNIX vendors included. One of our customers (I cannot say who they are) used the phrase 'authentication plumbing' when they described kerberos - I agree with this as it is a nice way to picture the Kerberos protocol in relation to other methods of authentication. What they were refering to is the unique and very capable ability of the Kerberos protocol to pass credentials across the network in a secure manner so that applications and services/devices know who the user/initiator is and also allow the service to delegate the credentials to another application component - ideal in an n-tier architecure which is very common in todays application solutions. If you combine this with PK and other forms of authentication then you can take advantage of private keys on client workstations without having the user to authetnicate more than once and hence get closer to the dream of 'secure single signon'. I can also introduce the web authentication environment and it is clear that this is where PK has traditionally been used, but in my experience most implementations use a web server certificate only for SSL purposes (not a client browser certificate held in a secure device such as a smart card) - hardly as secure as having a unique session key for each authentication request. I know that you will see more use of PK and Kerberos together in a web environment. One such example is that Kerberos is a good solution for passing credentials from browser to web server, then have the web server delegate the credentials to another application component such as an application server and then perhaps onto a database server ... The delegation of credentials cannot easily be acheived using PK technology - certainly not in a standard way anyway. I hope you have seen from my feedback above that there is clearly space for both a PKI and a KI (Kerberos Infrastructure) in many networks - even the Internet will need a global KI if Microsoft/IBM are successful with their WS-Secure proposal for Passport/WebServices and .NET authentication. I have no comments on KINK as I haven't studied it closely. You mentioned SSH in your posting - this is a protocol that is widely used, but still needs key management and improved authentication for operating system access. I can explain this better if you can contact me. I also noticed that you prefered an opensource solution ? Is this because of cost ? Once again - I am sorry this post is so long. I also hope that you find my feedback useful. If you have any further questions or comments I would welcome continued discussion on this matter. Cheers, Tim. ________________________________________________ Kerberos mailing list [EMAIL PROTECTED] http://mailman.mit.edu/mailman/listinfo/kerberos
