Lightweight Directory Access Protocol (LDAP) Proxied Authorization Control
draft-weltman-ldapv3-proxy-13

[Ballot discuss]
This document says:
  >
  > The criticality MUST be present and MUST be TRUE. This requirement
  > protects clients from submitting a request that is executed with an
  > unintended authorization identity.
  >
  > Clients MUST include the criticality flag and MUST set it to TRUE.
  > Servers MUST reject any request containing a Proxy Authorization
  > Control without a criticality flag or with the flag set to FALSE with
  > a protocolError error.  These requirements protect clients from
  > submitting a request that is executed with an unintended
  > authorization identity.
  >
  I think the author intended to delete the first paragraph when the
  second one was added to resolve the concerns raised bu Ned Freed.

[Note]: 'Kurt to follow up with auth and get revised text; he will join as co-author by Jan 1. 2005 if no revision seen by them.' added by Ted Hardie

[Ballot comment]
I'm very confused by this document -- what, precisely, does it define?
Is its whole purpose the definition of the OID?  Proxies done properly
are typically bound to the principal to whom some access right is being
delegated, and often contain a set of restrictions as well.  This seems
to give carte blanche to anyone who has the magic string to do anything
at all.

[Ballot comment]
Mmmm.. I see reference:

  [KEYWORDS] Bradner, Scott, "Key Words for use in RFCs to Indicate
        Requirement Levels", draft-bradner-key-words-03.txt, January,
        1997.

I guess that should just be RFC2119 !!

[Ballot comment]
My no-objection is based on a lack of time to review the preceding work on LDAP
authentication and authorization that seems to be presumed here.  It would be a
good idea if there some delineation of these dependencies, even if brief.  (I'm
not requiring a change, must making the comment).

[Ballot comment]
Spencer Dawkins, Gen-ART reviewer, has made some editorial comments.
To save space on the evaluation form, I've recorded these in the tracker log.
They are also at

Comments from Spencer Dawkins, Gen-ART reviewer

3. Proxy Authorization Control

  A single Proxy Authorization Control may be included in any search, compare, modify, add, delete, modify DN or extended operation request message with the exception of any extension that causes a change in  authentication, authorization, or data confidentiality [RFC 2829], such as Start TLS [LDAPTLS] as part of the controls field of the LDAPMessage, as defined in [RFC 2251].

Spencer: I know it's an editorial nit, but "as part of" completes a phrase that left off SIX COMMAS earlier. Suggest alternative text as

"A single Proxy Authorization Control may be included in any search, compare, modify, add, delete, modify DN or extended operation request message as part of the controls field of the LDAPMessage, as defined
in [RFC 2251], with the exception of any extension that causes a change in authentication, authorization, or data confidentiality [RFC 2829], such as Start TLS [LDAPTLS]."
...

  The controlValue SHALL be present and contain either an authzId [AUTH] representing the authorization identity for the request or

Spencer: "or SHALL be empty..."

  empty if an anonymous association is to be used.

Spencer: "anonymous authorization identity"?
...

  If the requested authorization identity is recognized by the server, and the client is authorized to adopt the requested authorization identity, the request will be executed as if submitted by the proxy authorization identity, otherwise the result code TBD is returned.

Spencer: I'm new to the world of IANA LDAP result codes, but wouldn't you expect to see a proposed name for this result code in this document?
Also in IANA Considerations section later on...

  [Note to the IESG/IANA/RFC Editor: the value TBD is to be replaced with an IANA assigned LDAP Result Code (see RFC 3383 section 3.6]
...

5. Security Considerations
...

  Note that the server is responsible for determining if a proxy authorization request is to be honored. "Anonymous" users SHOULD NOT be allowed to assume the identity of others.

Spencer: It seems like this restriction should appear a lot earlier in the specification - it's an implementation consideration, not ("just")
a security consideration...
...

8. Normative References

  [KEYWORDS] Bradner, Scott, "Key Words for use in RFCs to Indicate Requirement Levels", draft-bradner-key-words-03.txt, January, 1997.

Spencer: Scott says this wins the plaid bunny award for most outdated reference in an ID this year - should be "RFC 2119", right?

[Ballot discuss]
I'm very confused by this document -- what, precisely, does it define? 
Is its whole purpose the definition of the OID?  Proxies done properly
are typically bound to the principal to whom some access right is being
delegated, and often contain a set of restrictions as well.  This seems
to give carte blanche to anyone who has the magic string to do anything
at all.  The security considerations section doesn't even mandate that
this magic string be kept confidential -- and its more powerful than
plaintext passwords, which we bar.

[Ballot discuss]
This document says:
  >
  > The criticality MUST be present and MUST be TRUE. This requirement
  > protects clients from submitting a request that is executed with an
  > unintended authorization identity.
  >
  I agree with all of the comments made by Ned Freed.  Specifying
  the behavior of both the client and the server is very important.
  Ned suggests the following:
  >
  > Clients MUST include the criticality flag and MUST set it to TRUE.
  > Servers MUST reject any request containing a Proxy Authorization
  > Control without a criticality flag or with the flag set to FALSE
  > with a {?} error.  These requirements protects clients from
  > submitting a request that is executed with an unintended
  > authorization identity.

[Ballot comment]
No one is happy with the current situation regarding access control
  in LDAP.  Leaving the determination of proxy access rights as an
  implementation detail is not going to improve the situation, and it
  will probably make it worse.  However, I have no guidance to offer.

[Ballot discuss]
This document says:

  The criticality MUST be present and MUST be TRUE. This requirement
  protects clients from submitting a request that is executed with an
  unintended authorization identity.
   
Client requirements like this only protect clients that follow the rules.
A server requirement, on the other hand, protects against broken clients.
Given the likihood of security violations and screwy behavior if this
isn't enforced, I suggest changing this to read something like this:

  Clients MUST include the criticality flag and MUST set it to TRUE.
  Servers MUST reject any request containing a Proxy Authorization
  Control without a criticality flag or with the flag set to FALSE
  with a  error. These requirements protects clients from
  submitting a request that is executed with an unintended
  authorization identity.

I'm not wild about leaving the determination of proxy access rights
as an implementation detail, but given the situation surrounding
access control in LDAP in general I suppose we can do no better
at this time.

Nit: (date) in copyright field

[Ballot discuss]
This document says:

  The criticality MUST be present and MUST be TRUE. This requirement
  protects clients from submitting a request that is executed with an
  unintended authorization identity.
   
Client requirements like this only protect clients that follow the rules.
A server requirement, on the other hand, protects against broken clients.
Given the likihood of security violations and screwy behavior if this
isn't enforced, I suggest changing this to read something like this:

  Clients MUST include the criticality flag and MUST set it to TRUE.
  Servers MUST reject any request containing a Proxy Authorization
  Control without a criticality flag or with the flag set to FALSE
  with a  error. These requirements protects clients from
  submitting a request that is executed with an unintended
  authorization identity.

I'm not wild about leaving the determination of proxy access rights
as an implementation detail, but given the situation surrounding
access control in LDAP in general I suppose we can do no better
at this time.

Nit: (date) in copyright field