Re: [ldapext] draft-zeilenga-ldap-dontusecopy-00.txt

["Kurt D. Zeilenga" <Kurt@OpenLDAP.org>]

At 12:25 PM 6/23/2005, Jim Sermersheim wrote:
>I'm struggling to understand what is meant by the terms "copy" and "authoritative". I don't see one as the antithesis of the other.

I am using the term 'authoritative' in the same manner which
X.500(93) uses the term 'original'. I am using the term
'copy' in the same manner as the X.500(93) term 'copy'.

>How is a "copy" distinguished from an "original"?

  Replication in the Directory refers to the existence of
  copies of directory entry information and operational
  information held by DSAs other than the DSA responsible
  for the creation and update of the information. This DSA,
  containing the original information, is called the master
  DSA.
...
  A DSA holding copied information forwards all requests that
  would modify the copy information, and all requests that
  indicate that copy information shall not be used, to the
  master DSA holding the information.

>How does the term "copy" relate to the term "shadow"?

"Shadow" is an adjective that applies to some other term,
such as "copy".  A "shadow copy" is information copied
through shadowing.

>How do these terms relate to the term "read-only"?

In X.500, copies are always read-only.

>To me, the term "authoritative" is completely orthogonal to most of this. While an authoritative entry is probably never read-only, the determination of the placement of an authoritative entry may be completely arbitrary or application specific. For example, an address book application may consider a group of entries held on DSA1 to be authoritative, while a provisioning application may consider many of those same entries to be authoritative only when accessed on another DSA. Other applications may wish to share the notion of where an entry is said to be authoritative.

I'll replace "authoritative" with "original" to align with X.500(93).

>In my experience, the need for a control which specifies an operation take place on the authoritative copy typically has to do with update operations. 

As update operations can never be applied to a copy, there is
no need to have a control which says the update operation must
be applied to the original.

>One DSA is said to be authoritative for an entry,  and thus update operations may be asked to be directed at that authoritative copy. This is obviously especially true in multi-master systems.

In X.500(93), only one DSA can be the master for any particular entry.

>If this is what the I-D seeks to provide, I suggest it rename the control to "useAuthoritative", and allow it to be applied to all operations (rather than interrogation only). The problem is in defining how to determine whether a given entry is authoritative.

not.

>If, on the other hand, it seeks to force an interrogation operation to be directed to a non-read-only, or non-shadow copy of an entry, it should not introduce the term "authoritative". Rather it should follow X.500 and replace that word with "non-shadow"

Yes.  I will use the term 'original'.

(Of course, X.500 use of the term 'master copy' is confuses
matters.  Which, BTW, is why I introduced the term 'authoritative'.)

Kurt

> 
> 
>Jim
>_______________________________________________
>Ldapext mailing list
>Ldapext@ietf.org
>https://www1.ietf.org/mailman/listinfo/ldapext


_______________________________________________
Ldapext mailing list
Ldapext@ietf.org
https://www1.ietf.org/mailman/listinfo/ldapext