RE: LDAPsubentry

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

At 11:06 AM 11/19/99 +0100, Erik Andersen wrote:
>The consequence of making it an abstract object class would be that the 
>different types of subentries each should have a structural object class 
>defined instead of an auxiliary object class (as we have it X.500).

No.  They could share a structural class subclassed from the
abstract class to provide cn naming.

Note: the intention of the LDAP subentry was for other subentries
to subclass ldapSubentry.  Your statement implies that the author
may want to rework the draft to indicate that implementors should
mix in auxiliary object classes to extend ldapSubentry for
specific uses.

>A naming attribute would have to be defined for each of those.

Yes.  But they MAY also share one structural object class.  However,
they would be also allowed to separately subclass the abstract class
if more appropriate for their application.

>This naming attribute 
>could then be different for each type of subentry. What would the advantage be 
>of that?

Some subentries can consume a huge amount of namespace.  If they
are forced to share 'cn' naming with other subentries, name space
conflicts will arise.  Such subentry need to use a naming attribute
which is not used otherwise to ensure no conflicts arise.

The benefit of defining both an abstract AND a structural
object class is that it offers flexibility at no significant cost.

MUST 'cn' will force those who cannot provide a cn attribute
to define other objectclasses for subentries which are not
subclasses of LDAPsubentry.  This would be bad.

>As the same clients will access X.500 systems and LDAP servers, it must be an 
>advantage for implementers not to have unnecessary differences.

I think LDAPsubentry differs enough already that X.500 clients are
likely to be quite lost already.  Dividing the class into two
(an abstract and a structural) object class would likely not
cause additional heartburn.