W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > January to April 1998

Re: More comments on draft-ietf-http-authentication-01.txt

From: Life is hard... and then you die. <Ronald.Tschalaer@psi.ch>
Date: Mon, 6 Apr 1998 02:00:25 +0200
Message-Id: <98040602002493@psicla.psi.ch>
To: GISLE@aas.no, HTTP-WG@cuckoo.hpl.hp.com

[Gisle Aas wrote:]
> 1) It is not perfectly clear to me if the namespace of realms are
> separate for each authentication scheme or even separate when using
> the server as a proxy and when not.  I matters to me becase I try to
> index authentication objects in the client per "server/realm".
> 
> This sentence:
> 
>   The realm value (case-sensitive), in combination with the canonical
>   root URL (see section 5.1.2 of [2]) of the server being accessed,
>   defines the protection space.
> 
> make me believe that I can assume that it would not be possible to
> have to deal with all the following as separate "protection spaces"
> the same time:
> 
>    Authorization: Basic realm="foo"
>    Authorization: Digest realm="foo"
>    Proxy-Authorization: Basic realm="foo"
>    Proxy-Authorization: Digest realm="foo"

Hmm, I assumed the protection space to be defined by the tuple
(host, port, scheme, realm), but you are right that it's not clearly
specified. However, I would not differentiate between Authorization and
Proxy-Authorization, although at least one popular server will allow
different auth info to be set up for a proxy and server on the same
host and port.

> 2) How to interpret the URIs in the 'domain' attribute of the Digest
> WWW-Authenticate is not clear to me either.  Are the URIs in this
> list path prefixes that define protection spaces or must there be
> an exact match.  If it is a prefix, what happens if the URI contains a
> query part.

An exact match. Although in my implementation I actually use these to
create new protection space tuples (if the host or port is different)
and to update a cache of URL's. This cache is then used to guess if a
specific URL needs authentication (for preemptive auth header
generation). I think this fits the intent too.

> 3) Section 3.6 does not seem to know about the "407 Proxy
> Authentication Required" status code.  I though that a
> "Proxy-Authenticate" header could only be present in a 407 response.
> That is at least how I read <draft-ietf-http-v11-spec-rev-03>.  I
> would like this paragraph to go away (it would complicate my
> implementation if it stays):
> 
>   Note that in principle a client could be asked to authenticate
>   itself to both a proxy and an end-server. It might receive an
>   "HTTP/1.1 401 Unauthorized" header followed by both a WWW-
>   Authenticate and a Proxy-Authenticate header. However, it can
>   never receive more than one Proxy-Authenticate header since such
>   headers are only for immediate connections and must not be passed
>   on by proxies. If the client receives both headers, it must
>   respond with both the Authorization and Proxy-Authorization
>   headers as described above, which will likely involve different
>   combinations of username, password, nonce, etc.

How come this would complicate the implementation? Instead of doing

  if (status == 401)
      handle_auth_header
  if (status == 407)
      handle_proxy-auth_header

it becomes

  if (status == 401)
      handle_auth_header
      handle_proxy-auth_header
  if (status == 407)
      handle_proxy-auth_header

Actually, now that you mention it, I'm not even really sure why the 407
is needed at all. The Authorization vs. Proxy-Authorization header gives
you enough info on where the challenge is coming from. Can anybody shed
some light on why 407 exists?

> 4) Are there any servers that already implemented that that knows
> about "MD5-sess" and "auth-int" I can test against?

I waiting too...

> For those that can read Perl, I even include the source for my
> implementation here.  This code might even be of use to others
> implementing Digest or perhaps somebody can tell me if I misunderstood
> something in the spec by looking through the code.

I think there is a problem with the nonce-count and the MD5-sess
implementation. Assuming that _set_authorization is only called upon
receipt of a Authentication or Proxy-Authentication header (i.e. it's
not used for preemptively sending auth info) you need the following
changes:

> sub _set_authorization
> {
[snip]
>     my $uri = $req->url->full_path;
>     my $nonce = $self->{nonce};  $nonce = "" unless defined $nonce;
>     my $nc = sprintf "%08x", ++$self->{nonce_count};

You need to reset the counter when a new nonce is received. Something
like

      if ($self->{prev_nonce} ne $self->{nonce}) { $self->{nonce_count} = 0; }
      $self->{prev_nonce} = $self->{nonce};
      my $nc = sprintf "%08x", ++$self->{nonce_count};

>     my $cnonce = sprintf "%x", rand(0x1000000);
> 
>     my $a1;
>     if ($algorithm eq "md5") {
> 	# A1 = unq(username-value) ":" unq(realm-value) ":" passwd
> 	$a1 = MD5->hexhash("$user:$realm:$pass");
>     } elsif ($algorithm eq "md5-sess") {
> 	# The following A1 value should only be computed once
> 	$a1 = $self->{'A1'};
> 	unless ($a1) {

should be

	unless ($a1  &&  ($self->{stale} eq "true"))

(I believe this will be clarified in the next draft)

> 	    # A1 = H( unq(username-value) ":" unq(realm-value) ":" passwd )
> 	    #         ":" unq(nonce-value) ":" unq(cnonce-value)
> 	    $a1 = MD5->hexhash("$user:$realm:$pass") . ":$nonce:$cnonce";
> 	    $a1 = MD5->hexhash($a1);
> 	    $self->{'A1'} = $a1;
> 	}
>     } else {
> 	return;
>     }
[snip]


  Cheers,

  Ronald
Received on Sunday, 5 April 1998 17:09:19 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:33:14 EDT