Re: Unique identifiers for LDAP attributes

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

At 03:02 PM 7/13/00 +0100, David Chadwick wrote:
>However, we have the situation that some LDAP servers do not 
>require OIDs to be defined for attribute types,

Which implies they cannot properly publish schema...
Which implies they must be read-only servers...

>and the LDAP spec 
>deprecates the use of OIDs in protocol in preference to strings.

RFC2251 recognizes that names are non-unique but requires servers
to use them.  This does seem quite odd.

>Given that many LDAP clients now map the attribute type strings 
>from protocol into a user friendly language dependent display string, 
>the string representation in protocol has about had its day and 
>served its purpose. Isnt it about time that we altered the LDAP 
>spec to recommend that OIDs be the preferred way of transferring 
>attribute types in protocol, and that the OIDs become the globally 
>unique way of identifying attribute types.

Maybe then clients would actually discover (and make use) of
published schema...

I would support lifting the MUST use short names requirement.
This requirement is not needed to support interoperability and
hence, per RFC2119, it shouldn't be a MUST.

I would support stating that servers MUST use a non-ambiguous
identifier.  That is, they must either ensure that NAME of given
schema elements are non-ambiguous (with a subschema subentry)
or use OIDs.