[Date Prev][Date Next] [Chronological] [Thread] [Top]

RE: Unrecognized critical control & referrals



I've always assumed that a server which recongizes a control
supports that control.  Note that the RFC2251 never uses
the term "supported" in the context of controls though it
does state to return the unsupportedCriticalExtension
(unavailableCriticalExtension [12]) resultCode.

Anyways, I agree that the server should return
unsupportedCriticalExtension (unavailableCriticalExtension [12])
in this case.

	Kurt

At 10:29 AM 6/14/00 -0400, Salter, Thomas A wrote:
>The server can only do that if it knows something about the control.  If the
>server can somehow tell that it would have returned a referral even if it
>supported the control, then it can go ahead and return the referral.
>
>For instance, you could build a server that recognizes the sort control, but
>does not support it.  It could safely return a referral (or even a single
>result), but would need to return unsupportedCriticalExtension if the search
>returned two or more entries.
>
>In the general case the server can't guess about nature of the control.  For
>all it knows, the control has some obscure meaning such as "under no
>circumstances return a referral for this request".
>
>I think the key words in the RFC are correct: "... does not recognize ...".
>This allows a server to recognize, but not support, a control.
>
> > -----Original Message-----
> > From: Kurt D. Zeilenga [mailto:Kurt@OpenLDAP.org]
> > Sent: Wednesday, June 14, 2000 9:17 AM
> > To: ietf-ldapext@netscape.com
> > Subject: Unrecognized critical control & referrals
> > 
> > 
> > RFC 2251, 4.1.2:
> >    If the server does not recognize the control type and the 
> > criticality
> >    field is TRUE, the server MUST NOT perform the operation, and MUST
> >    instead return the resultCode unsupportedCriticalExtension.
> > 
> > There seems to me the need for an exception to the MUST to allow
> > a server to return a referral instead of 
> > unsupportedCriticalExtension.
> > If the request would, if the control was not marked critical,
> > have returned a referral result, shouldn't that same referral be
> > returned when the control is critical?
> > 
> > Kurt
> >  
> > 
>
>