Abandon'ing update operations (Was: WG Consensus: Abandon Operation)

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

I've changed the subject to distinguish them from the
issues we've reached consensus upon (as detailed in
the 'WG Consensus: Abandon Operation' post)...

At 12:51 PM 2002-08-09, David Chadwick wrote:
>I therefore think that further clarification text is needed in addition
>to what you have already proposed, so as to remove any ambiguity over
>the use of Abandon on updates. It seems like we have three choices
>
>i) To state explicitly that Abandon cannot be used against update
>operations, as per the X.500 protocols. We may need to state what
>happens if this is attempted (e.g. protocol error or similar)
>ii) To state explicitly that Abandon can be used against update
>operations, and to state what the outcomes are in all conceivable cases
>(new text will be needed here as X.500 wont give you any help)
>iii) To state that this document gives no guidance as to the use of
>Abandon with update operations, and it is a local implementation matter
>how they are handled.

In my opinion, RFC 2251 is fairly clear that:
 a) the Abandon operation can be used to request the abandonment
    of any outstanding operation,
 b) this Abandon request may not be honored.

but I also believe:
 c) the Abandon operation should only be used to abandon search
    operations (abandoning compare is pointless, abandoning
    other non-update operations, e.g. bind and starttls, is
    problematic, and abandoning update is also problematic).

To address this, I think [protocol] should state that abandon
should not be used where the client needs a clear indication
of whether abandonment occurred or not and, then, discuss
cases where a clear indication is (generally) needed.

That is, don't prohibit implementation of Abandon operation for
non-search, just provide guidance for its use.

Kurt