What ever happened to Tandem and NonStop OS 2080

kerberos started with userid-pbuttword authentication as an infrastructure function ... and then handed out tokens-credentials regarding what authentication had been done. some number of applications that previously required their own user-pbuttword ... got kerberized to honor kerberos tokens.

the original pk-init draft for kerberos added an option where public key could be registered in lieu of pbuttword and simple digital signature authentication be performed

w-o requiring PKIs, digital certificates, etc.

somebody (periodically sends email apoligzing for having done it), was responsible for getting the kerberos pk-init draft upgraded to also allow PKI operation and digital certificate operation.

the theoritical PKI design point is that somebody can submit a digital signature for authentication with an appended digital certificate for identification ... and the digital certificate would also specify the privileges granted the enbreasty ... aka w-o requiring the enbreasty to have been separately registered to any system (i.e. a perfect stranger, having presented a valid digital certificate would be allowed access to your system w-o any additionly processes).

what happens tho, is that typical systems still require some sort of independent registration of valid enbreasties .... in which case, it is perfectly acceptable to include the enbreasty's public key registration as part of the process ... making any appended digital certificate redundant and superfluous (as well as the PKI and buttociated certification process also redundant and superfluous).

fundamentally, the PKI design point was for first time communication between complete stangers ... that both parties then could rely on a trusted third party to provide introduction (aka the letters of credit-introduction from the sailing ship days). one of the reasons that for PKI being redundant and superfluous in environments which have pre-existing relationships, is that such pre-existing relationships negate the original design point for PKIs.

in any case, there have been a number of processes for doing the initial kerberos authentication as upgrades to the origina; userid-pbuttword paradigm ... certificateless digital signature, certificate-based digital signature, hardware token digital signature, other kinds of hardware token technology.

from 3-factor authentication

* something you have * soemthing you know * something you are

... the kerberos original userid-pbuttword is a "something you know" authentication. the hardware token based operations typically are technologies that attempt to buttert the person is in possession of a unique "something you have" hardware token. many of the digital signature based operations are either directly "something you have" authentication (i.e. business process that embodies a unique private key in a hardware token) ... or a simulated "something you have" authentication (i.e. business process that attempts to emulate in software a hardware container for unique private key).

many of the userid-pbuttword operations are also shared-secret

where the "something you know" pbuttword is registered at the authenticating enbreasty.

hardware tokens can be used to implement two-factor authentication, where the hardware token is certified to required a pin-pbuttword for correct operation ... aka the relying party infers from the operation of the hardware token that the correct pin-pbuttword was entered ... w-o the relying party needing to know what the pin-pbuttword is.

one might claim that one of the reasons for ssh (sftp, scp, etc) uptake ... is that it continued to support the trivial certificateless public key registration and digital signature authentication ... while kerberos pk-init got all bollixed up requiring PKIs, digital certificates, etc ... as part of basic (redundant and superfluous) operation ... over and above also typically requiring the user to be registered.

--