W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > January to April 1995

Re: Digest Access Authentication proposal - "Authorization"

From: Owen Rees <rtor@ansa.co.uk>
Date: Wed, 22 Mar 1995 16:25:50 +0000
Message-Id: <9503221625.AA01613@plato.ansa.co.uk>
To: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
zurko@osf.org (Mary Ellen Zurko) writes:
> It seems that from the start of WWW security, the terms authentication
> and authorization have been used in ways that obfuscate the difference
> between them.
[etc]

I have just read through both section 10 of draft-ietf-http-v10-spec-00 and 
draft-ietf-http-digest-aa-00 and have found only one place where I think that 
either term is used incorrectly. This is under <domain> in 2.1 of digest-aa, 
where I think "authorization" should replace "authentication.

The credentials are authorization information; "This request is from <username>
 who claims the right to access <requested-uri> in <realm>" the digest is the 
authentication information for the claim "Only <username> could and would have 
created this <digest> of that claim and the secret we share". The server could 
have constructed the credentials, but this scheme is not trying to address the 
non-repudiation problem.

Reading through digest-aa-00 again has revealed two other problems:

1) <digest> and <message-digest> have only the nonce to link them in a 
request. Specifically, if the server re-uses nonces, the client cannot ensure 
that the message body matches the requested uri - this could be a problem 
POSTing.

2) There is nothing except the nonce to link the response to the request. If 
the nonce is re-used, an intruder could substitute an old response.

If there is any possibility that the client has used the same method:URI pair 
before, the only way to ensure that the entities in request and response 
messages are current is for the client to create a nonce of its own. This has 
to be added to both <digest> amd <message-digest> (both ways) and therefore be 
another parameter to the Authorization header so that the server can create 
the <message-digest> for the response. Guaranteeing a current response from 
the server only matters if the resource could change. If the client is happy 
with a cached version, then it could use method:URI in place of the nonce to 
ensure that the reply is to some version of this question.

Regards,
  Owen Rees <rtor@ansa.co.uk>
Information about ANSA is at <URL:http://www.ansa.co.uk/>.
Received on Wednesday, 22 March 1995 08:34:22 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:31:14 EDT