I have some contrary thoughts on the AAA section of this document.

Section 4.1 tries to describe requirements wherein the I2RS Clients may
request for subsets of AAA policy to be exported to the Client so that the
client may enforce them.  While this seems like a nice way to scale the
operations, in some cases disclosing those policies (even if we find a good
way to encode the AAA validation in a generic enough way to distribute) may
accidentally disclose information that is otherwise intended to be secure.

I would seek comment from the security directorate, but I suspect we don't
want to do this.

But in section 4.4, we try to discuss availability.  The first sentence
immediately says "enforcement should not remain local", while one way to
enable security in some environments is to distribute and synchronize policy
to be enforced locally.  

It then goes on to talk about general availability mechanisms and then we
further dive into security against DoS.

I believe we may be boiling the ocean a bit to try to go into too many
details about the design of secure AAA systems.  It seems a bit out of scope
for I2RS to do such work; we should defer to work done elsewhere on the
topic, if it exists.  If it doesn't exist, I'm not sure we should do it.

What is right for us to point out is, "If we use a remote AAA mechanism, it
must be robust in hostile environments".  Expand that as you will, but being
too proscriptive is not our job.

-- Jeff

_______________________________________________
i2rs mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/i2rs

Reply via email to