RE: current control combination proposals

["Ramsay, Ron" <Ron.Ramsay@ca.com>]

constrainViolation is only appropriate for update operations.

-----Original Message-----
From: owner-ietf-ldapbis@OpenLDAP.org
[mailto:owner-ietf-ldapbis@OpenLDAP.org]On Behalf Of Chris Ridd
Sent: Sunday, 16 May 2004 01:01
To: Kurt D. Zeilenga; Jim Sermersheim
Cc: ietf-ldapbis@OpenLDAP.org
Subject: Re: current control combination proposals


On 14/5/04 11:36 pm, Kurt D. Zeilenga <Kurt@OpenLDAP.org> wrote:

> At 02:48 PM 5/14/2004, Jim Sermersheim wrote:
>> I'm having a hard time gauging consensus on this. Maybe distilling it to
>> the core changes will help:
> 
> I think it would also help if we surveyed developers of existing
> implementations to ask them how they handle various combinations
> of controls.

Not sure I want to contribute to this thread, but here goes anyway...

Since controls are sent ordered, ie in a SEQUENCE OF, surely we need to
discuss the *permutations* of controls rather than the combinations? In
other words, does a server allow operation X with controls A+B, or with
controls B+A?

If incompatible controls are sent, but one is marked critical and one isn't,
the ordering of the encoded controls may well be significant.

I'd also suggest that an error code that's perhaps appropriate for
signalling "incompatible" controls would be constraintViolation. Sending a
protocolError seems a bit harsh.

Cheers,

Chris