[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: New text for X.501


Ron,

Ramsay, Ron wrote:
I must protest that DirectoryString is one of the cases where the 
> abstract value can be preserved. In order to break it you have had
> to go X.500 -> LDAP -> X.500 and I don't see how this can be justified.
> LDAP is an access protocol, so we should never have the case to which
> you referred.

This transformation occurs if an X.500 DSA uses LDIF to dump and reload
data. It also occurs if an X.500 DSA uses LDAP to chain to a subordinate
(LDAP-only) DSA.

On another matter, does PostalAddress allow spaces either side of the 
> '$'? If it doesn't, then this also is a syntax for which values can be
> preserved.

The problem with PostalAddress is the need to escape any '$' character
in the abstract value. The escape sequence is case insensitive and could
be case inverted by LDAP -> BER -> LDAP.

Also, relating to the syntaxes that cannot be preserved, your previous 
> example seemed to concern configuration items like schema. While it may
> be appropriate to configure LDAP directories using LDAP, I don't know
> that this should be the case with X.500 directories.

It's a necessity. There are LDAP applications that configure the schema
they need by using LDAP to modify the subschema subentry. Our directory
server let's them do it.

Regards,
Steven

> However, this still leaves items like SearchGuide. 

Ron

-----Original Message-----
From: owner-ietf-ldapbis@OpenLDAP.org
[mailto:owner-ietf-ldapbis@OpenLDAP.org]On Behalf Of Steven Legg
Sent: Tuesday, 28 October 2003 14:03
To: David Chadwick
Cc: LDAP BIS
Subject: Re: New text for X.501



David,

David Chadwick wrote:

Steven

We are talking about the abstract values and not the encoded values.
X.500 does not mandate any particular ASN.1 encoding rules except in the
case of the X.509 attributes which must be DER (and even this was later
realised to be a bug, but it is too late to change it now). DSAs/LDAP
servers will typically store local representations of abstract values
and what these are is not stated, its implementation dependent (although
in some cases such as the X.509 attributes, they will have to store the
encoded values so as to preserve the signatures). Whether the transfer
syntax is BER, LDAP or XER should be irrelevant to the abstract value.
Does this help to clarify what the intention is



Yes, though you should be aware that there are cases where even the
abstract value can't be preserved. The chosen alternative in a DirectoryString
isn't guaranteed to be preserved in a transformation from BER to LDAP-specific
to BER. This affects the Directory String syntax, as well as various syntaxes
that have embedded DirectoryString components. The chosen alternative is
inconsequential for directory purposes, but changing it is, by strict definition,
a change to the abstract value.

Regards,
Steven


regards

David


Steven Legg wrote:


David,

David Chadwick wrote:


Dear LDAPers

At the recent Geneva meeting of the X.500 group, Defect Report 303 was
discussed. This concerns the fact that a user cannot be guaranteed that
the information presented to LDAP/X.500 server in an update operation is
subsequently returned unaltered in a Search operation. Due to this, in
the PKIX work we are adding text to the IDs specifically to say that for
X.509 certificates and CRLs the data must not be altered by the LDAP
server. The X.500 group is going to go one step further than this and
state that no attributes must be altered by the server and must be
returned exactly as presented,


I think this change is ill-advised, as the requirement cannot be enforced
in a mixed LDAP/X.500 distributed environment. An attribute value that is
entered in an LDAP-specific encoding has to be transformed into BER to be
carried in DSP or DISP. There is no guarantee that the exact LDAP-specific
encoding of the original attribute value will be reconstructed by the
receiving DSA. Preservation of the exact encoding of PKI attributes can
only be made to work in the general case because the LDAP encoding and the
X.500 encoding is the same - BER. For most syntaxes this is not the case.

The XED specifications introduce a third way of encoding directory data (DXER),
which only increases the difficulty of preserving the original encoding.

If the X.500 standards add this requirement then, as a practical necessity,
I will have to disregard it. However, I will continue to preserve the abstract
value of attribute values to the extent that it is possible to do so.

Regards,
Steven



although a server may store a
canonicalised form for efficient matching if it so desires.

The defect report can only address the 1997 and 2001 versions of X.500,
since the 1993 version that LDAP is based in is no longer supported by
ITU-T/ISO.

Here is the gist of the proposed text to fix the defect report.

Stored attribute values must be held as supplied. We propose to add text
to X.501 in clause 8.5 and in 8.8.1, where we will point out that
rationalizations to stored values for the purposes of matching do not
effect the stored value. We will also add text to clause 6.1 of x.520
stating that the rationalizations describe in the matching rules are
ephemeral, for the purpose of the match only, and will not affect the
stored value.

Regards

David