[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: XACML OID tag?



Some responses to Bob's points and others.

1) There is no OID and if there was I have no idea what the semantics of the
OID would be.

2) The way to link a SAML assertion to a certificate is to use the KeyInfo
element of the SubjectConfirmation field to identify the authentication
credentials of the subject to which the assertion is bound. This may be
either by the Certificate itself, by the certificate serial and issuer or by
the public key parameters.

3) Although SAML attribute assertions are similar to X509.v3 attribute
certificates there are significant differences in the models. With attribute
certificates the model is of issuing static pieces of data that have long
validity intervals (days at least). The attribute certificates require the
issuer to establish parallel identity and attribute certification
hierarchies. The idea that end entity clients would ever implement cross
certification is bogus. The idea that end entities are going to implement a
functioning bifurcated authentication and attribute cross certified
heterarchy with separate roots of trust is bogus with knobs on.

4) For the specific SAML uses of attribute assertions it is the PDP that
performs the attribute processing and not the enforcement point (i.e. the
device hooking into the system). The important upshot of this is that it is
not necessary to standardize implementation of policy languages in the end
entity devices.

5) SAML does not provide the status management functions of X.509 but the
SAML framework may be readilly extended to support them, not least because
XTASS, the original assertion framework on which SAML is based supported an
assertion status mechanism (tier 4).

If an attribute statement is simply a statement about a subject the chances
are that a realtime OCSP type system will be much more satisfactory than a
static certificate based system. The number of accesses within the validity
interval, the sparseness of the data, etc. is simply to small to make
certificate issue worthwhile. 

The intended use for the tier 4 meta-assertion scheme is to manage
assertions that map to negotiable financial instruments.

6) I agree with Hal on the points he made, however remember that the SAML
working group has a very narrow scope. The assertion framework on which SAML
is based was designed as a very general purpose vehicle.

7) We examnined the use of XER encoding at length for a full 6 minutes early
in the design of XTASS/XKMS/SAML/XKASS/XTAML when it was all called TAXI
(Trust Assertion XML Infrastructure). It would be a completely worthless
exercise. There is no intrinsic value in the XML syntax over the X.509 great
enough to outweigh the fact that there is a massive X.509v3/BER installed
base. There is no value therefore in introducing a new PKI syntax unless the
semantics are also changed to allow uses that were not possible before.

8) The idea of canonicalization is broken. It was broken in DER it is still
broken in XML Signature. If you want to be able to check signatures put the
responsibility on the infrastructure not to mess with the bits. The minute
you try to support applications that slightly mangle data you lose big. The
ability to burst a certificate into components and put it in a directory
*and then delete the original* is bogus, was bogus and will remain bogus.

If digital signatures had been implemented in the very first email systems
then gateways would never have got into the idiotic charater set and line
wrapping transformations that break so many messages. That task should
always have been performed by the end client. The reason IP works and many
competitors do not, is that unlike other schemes IP does not mangle the
contents of each packet in transit.

9) If you really wanted to you could use X.509 as a signature envelope for
SAML assertions instead of XML DigSig (which is by the way a joint IETF-W3C
WG). But CMS would appear to be a much easier and simpler solution.
Alternatively you could profile XML-Dig sig, use the minimal c14n etc. to
make implementation trivial.


	Phill



Phillip Hallam-Baker FBCS C.Eng.
Principal Scientist
VeriSign Inc.
pbaker@xxxxxxxxxxxx
781 245 6996 x227

 

Attachment: Phillip Hallam-Baker (E-mail).vcf
Description: Binary data