RE: WG last call on "duplicate entries"

[Dave Watts <DJW@dataconnection.com>]

Kurt,

I agree that we should clarify RFC2891, and then make the dupent I-D
consistent with that.

In terms of how we should clarify it, we disagree. I think it's more natural
from a user point of view (though in general harder to implement!) that
sorting should encompass subtypes. If sorting (and hence dupent) does
include subtypes, then we further disagree on what this actually means.

My opinion isn't particularly strong on sorting and subtypes - it's more
important to agree on something which is well-defined.

Cheers, 

David


-----Original Message-----
From: Kurt D. Zeilenga [mailto:Kurt@OpenLDAP.org]
Sent: 28 March 2001 19:02
To: Dave Watts
Cc: 'jimse@novell.com'; 'roland@catalogix.se'; ietf-ldapext@netscape.com
Subject: RE: WG last call on "duplicate entries"


At 04:29 PM 3/27/01 +0100, Dave Watts wrote:
>... I'd treat values of 'cn' and 'sn' as if they were all values of the
same
>att 'name', and thus expect 3 entries to be returned:

But's not how subtypes are treated in general.  When you do a
search and ask for 'name' back.  A separate attribute for name
and each subtype of 'name' is returned.

>I think this aligns better with how values would be sorted (although
RFC2891
>doesn't explicitly talk about subtyping at all).

I think without clarification to RFC 2891, it's unclear to how
any dupent subtype handling aligns with RFC 2891.  However,
as this control is specifically designed to (optionally)
be used in conjunction with a sort control, I think we should
first discuss on RFC 2891 treats subtypes.

The question, I guess, is whether or not it makes sense for
a single sort key to encompass multiple attribute types
(though subtyping).  My initial opinion is that it doesn't.
If this is the consensus view, it may actually be best to
revert the behavior of dupent so that it only duplicates
explicitly specified attribute types.  Of course, this could
be optional behavior (yet another BOOLEAN), as the
subtyping I describe (which I believe is consistent with
the I-D) is also useful.