Re: (Dumb?) X-ORDERED 'VALUES' question

[Howard Chu <hyc@symas.com>]

Quanah Gibson-Mount wrote:
> --On Friday, August 24, 2007 5:03 PM +0200 Pierangelo Masarati 
> <ando@sys-net.it> wrote:
>
> > I'm implementing the Control syntax in draft-chu-ldap-logschema, which
> > is used for the reqControls and reqRespControls as X-ORDERED 'VALUES',
> > and I'm having a little bit of trouble in understanding how I should
> > encode values.  Right now, I'm putting the DER encoded controls in the
> > pretty values, and the control OID in the normalized values as per the
> > objectIdentifierFirstComponentMatch equality matching rule.  However,
> > it's not clear whether I should put the ordering prefix in either value.
> >   If I do, matching fails because test_filter() ignores the fact that an
> > attribute may be ordered; in fact, it uses value_match() instead of
> > ordered_value_match().  I suspect this is a bug, but I might be missing
> > something __really__ obvious...
>
> Maybe unrelated, but this reminds me of the issue I ran into with valsort 
> where adding ordering made it so that I couldn't search on the non-weighted 
> values in the attr if it gets indexed, essentially requiring there to be 
> two attributes for any weighted attribute if you needed indexing.  For 
> example "ou" and a private namespace "zimbraOU" or something.

I suspect the same issue will exist with ordered values, but with ordered 
values we can actually fix it since the schema for the attribute lets us know 
that special treatment is needed. With valsort, nothing outside of the overlay 
knows that anything special is going on.
--
  -- Howard Chu
  Chief Architect, Symas Corp.  http://www.symas.com
  Director, Highland Sun        http://highlandsun.com/hyc/
  Chief Architect, OpenLDAP     http://www.openldap.org/project/