Rep (4) : QUIPU vs X.500 (was: A tool for...)

[Ascan Woermann <[e-mail]>]

> The reason that this is not an issue is that RFC 1276 is currently the
> only viable way to run the high levels of the DIT.   The fix is
> academic.   

We provide our users with tools to automatically maintain their knowledge
of high levels of the DIT. They do this using DAP and they are independent of
the DSA they interrogate. They have the advantage that they restate the
knowledge in terms of the local DSA's knowledge representation capabilities.
Such tools can be extended to shadow entries as well.
One also can envisage out-of-band replication protocols, e.g. the NADF.
So I think there are cheaper and simpler ways of achieving the goals of 
RFC1276.

> The converse is not a problem - non-RFC 1276 DSAs can be plugged in as
> leaf DSAs in an RFC 1276 tree.   QUIPU supports this.

A leaf DSA still needs to be able to represent an RFC 1276 DSA and its 
knowledge in its local database. Relying only on a superior RFC 1276 DSA for 
any non-local query implies a potential bottleneck. 
Thus, RFC1276 needs to be extended in order to state what an RFC 1276 DSA's
knowledge is i.t.o. the references defined by the standard. I think we're
talking about a special-case of the non-specific subordinate reference.
Since this is a optional feature of the '88 standard, it has to be stated
clearly that for an '88 DSA to integrate cleanly in an RFC1276 tree even
as a leaf DSA, it needs to support these references.
RFC 1276 DSAs need to be cleaned up to correctly handle these references
even in referral mode - a mode that the RFC doesn't exclude - and the OIFP
tests have shown that there are problems in this area for most
implementations. In fact, it seems likely that loop detection will become a 
prerequisite feature for DSAs as long as older buggy DSAs are around.
In sum there clearly is room for discussion and improvement.

Ascan