RE: ;binary transfer of the binary syntax

["Steven Legg" <steven.legg@adacel.com.au>]

Kurt,

> -----Original Message-----
> From: owner-ietf-ldapbis@OpenLDAP.org
> [mailto:owner-ietf-ldapbis@OpenLDAP.org]On Behalf Of Kurt D. Zeilenga
> Sent: Tuesday, 19 September 2000 3:46
> To: hahnt@us.ibm.com
> Cc: ietf-ldapbis@OpenLDAP.org
> Subject: Re: ;binary transfer of the binary syntax
> 
> 
> At 07:02 AM 9/18/00 -0400, hahnt@us.ibm.com wrote:
> >My preference is that ";binary" is NOT required to be sent 
> (nor should it be expected).
> 
> I tend to agree.  However, compatibility with inetOrgPerson might
> be an issue...
> 
> >I would prefer to see "language" along the lines of: 
> 
> language tags are an extension, hence out of scope of this forum.
> 
> >whatever is sent in on the request is what will be returned 
> (like attribute type alias names).
> 
> What are alias names?  Each attribute type has an OID and one or
> more alternative names.  The alternative names can be viewed
> as "aliases" for the OID.  The specification states no requirement
> for any particular alternative name to be favored.  The implementation
> is free to return the attribute using any of these names regardless
> of how the attribute was requested.  Some servers might favor the
> requested name, others might do favor the first or the last, others
> might choose one at random.  It really doesn't matter which is
> returned as the client should know that they are all equivalent.

They should know, but typically they don't. Client implementors
take shortcuts. They don't bother checking the schema to see if the
attribute has been returned under one of the alternative names.
Therefore it is prudent for us as server implementors to return
attributes under the name that was requested by the user. Of course we
are out of luck if the user just requested all user attributes, i.e. "*".
This is one of those situations where using names instead of OIDs in
the protocol bites.

Regards,
Steven