Re: [ldapext] VLV: a particular target entry

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

At 03:16 PM 2002-10-03, David Boreham wrote:
>> If every entry in the client's current view has vanished, you would just
>> attempt a re-position based on their current requested offset. They
>> would end up in a place that may or may not be confusing.
>
>Indeed.

Or return an error.

>> In my mind, for this particular issue, there doesn't need to be any
>> change to the VLV protocol. Only clarifications and further notes to
>> implementers.
>
>My point was that there is no general solution to the
>dynamic list content problem,

I believe there are at least two general solutions to this
problem.
  1) a cookie (stateful) approach which can be used to position the
  view based upon previous list state and returned results,
  2) a stateless approach which positions the view relative
  to a particular object in the list.

Both solutions, where continuity cannot be provided, can detect
the lack of continuity.

>only various suggestions to make it not so bad,

I disagree.  Both of the above approaches not only handle
minor list changes but deal with some major list changes...
and, where they cannot deal with changes, they can detect
that they cannot and return an error.  This is significantly
better ignoring the problem.

>in the abstract, with no cited
>application where the problem has proven to arise in the
>first place.

(Based upon your description of the Sun/Netscape server)
Take your address book client, select a view against the
Sun/Netscape server, add one entry from the list before
the target, then page down.  The provided view will skip
over one entry. That's a problem, it's real.

Kurt

_______________________________________________
Ldapext mailing list
Ldapext@ietf.org
https://www1.ietf.org/mailman/listinfo/ldapext