Kerberized authorization service

Russ Allbery rra at stanford.edu
Mon Feb 11 12:58:00 EST 2008


Ken Hornstein <kenh at cmf.nrl.navy.mil> writes:

> Let's take two real world counter-examples: the AFS PTS server, and John
> Hascall's double-secret authz server (Stanford's wallet might be another
> example, but I don't know enough about it to say so let's not bring that
> into the mix right now).

The wallet is not an authz service.  It's a secret data management service
which lets you acquire various private keys, passwords, and similar secret
data upon proof of your identity.  It implements authorization checks, of
course, but does not provide them as a general service.

Stanford's authz story right now isn't particularly coherent, like most
sites.  We use a wide variety of different mechanisms (PTS groups, LDAP
entitlements, file-based Kerberos ACLs, AD groups, and other things)
depending on the application and they're not particularly well-integrated.

I should note, though, that in my experience with authorization, the
network protocol isn't really the hard problem.  It's usually not too
difficult to find some way to convert from one protocol to another or to
export authorization data into the form that a particular application
wants it.  The really hard part of authorization isn't communicating the
information; it's dealing with lifecycle.

There are a lot of lifecycle problems, but the one that best illustrates
the general problem in my experience is this:  If someone in your
organization changes roles, how does their authorization data get updated
across all of the systems and applications to which they have access?  If
your answer involves any manual process apart from the HR change, you
lose.  :)

Stanford currently very much loses on this, in a wide variety of ways.  We
really only have one authorization system that copes correctly with role
status changes (provided that it's used properly), and it only knows how
to talk to the financial system and isn't (currently) usable as a general
authorization solution.  There is some active work in the Internet2 arena
on this, but not to the point where I think people are deploying it.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>



More information about the Kerberos mailing list