[Date Prev][Date Next]
slapd (was: Enhancement to implement V3 Referral / SearchReferenceHandling)
Given that most of this content is not related to the
patch, I've changed the subject.
At 12:03 AM 5/26/00 +0400, Dmitry Kovalev wrote:
>Several questions about this patch and referral support in general.
>1) What functionality does exactly this patch add?
This patch extends the -lldap (the client SDK) referral chasing
to support LDAPv3 referrals.
>2) Is there some document (rfc or draft or whatever) it is based on?
No. The existing RFC (1823) and current API drafts do not
detail referral handling. See ldap_set_rebind_proc(3) for
previous documentation (this will have to updated per Steve's changes).
>3) Has it something to do with "draft-ietf-ldapext-namedref-xx.txt"
I found in OpenLDAP's docs subdirectory?
Not directly. This I-D is about in directory representation of
referrals, not about client chasing of returned referrals.
>Which of these are supported in OpenLDAP?
The -devel server all referral forms described by the named
>How are they supported (primarily - is there some frontend support, or it is all left to backends)?
Depends on the referral type. Generally speaking, superior referrals
are managed by the frontend and subordinate/search references are
managed by the backends.
>Is this a most recent document on this matter? If not - where is updated document?
There is not an updated version available (yet).
>I am going to implement all of these ways in my back-sql (since I need this functionality anyway), and willing to make it in conformance to standards,
Becareful with the term 'standard'. These docs are I-D, not
Standard Track RFCs (note all of LDAPv3 is currently "Proposed
Standard", not even "Draft Standard, and certainly not "Standard".
>if they exist at all... Also, consider developing of some frontend support for this, to make implementation in other backends simplier.
The slapd code to support named referrals has been "done" for quite
some time. See back-ldbm for an example of how to implement.
>4)Also, there is some statement about "referral" objectclass in that document, which constraints that those "ref" attributes only can be found in "referral" objects.
I think you misread the draft. The draft notes that this is only
one use of the ref attribute type and also notes that a named
referral must be of object class referral.
The draft may need clarification... feel free to make comments
to the authors or the IETF LDAPext WG. It's best to do so
before the document is submitted to the IESG for review as
a Proposed Standard.
>I'm not very clear about subclassing and things (BTW, where could I learn more about it?), so it seems to me that such constraint would render "ref" almost completely USELESS (since I cannot add it to any entry I want), especially for base search! If I am wrong - please explain, why...
See the LDAP FAQ, in particular Chadwicks online X.500 book.
>Anyway, in case of back-sql, there can be no subclassing at all, so I'm going to support "ref" automatically in any objectclass...
Don't do that. Only generate a referal if the object class is
>Again, brief examination of OpenLDAP sources makes me think that referrals are returned only if objectclass is "referral", and "base search" logic is absent.
The SLAPD generates both referrals and search references.
>Any information on these questions is appreciated, because it will save me lots of time looking into sources and RFCs and things... :)
You really need to read the RFCs and I-Ds AND look at the sources.
Of course, we're here to help, but you are expected to do your