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

draft minutes of April meeting



Here are the minutes I took for this meeting.  Sorry for the delay in posting
them to the list.


Draft minutes for the LDAPEXT Working Group meeting at Spring 1998 IETF

LDAPEXT met in three sessions on April 1st, 1998.

First Session

1. Agenda Review

Several additional topics were added to the agenda: Start TLS and 
recommended authentication methods, security parameters for signed 
operations, transactions, schema constraints and referrals.  

A request to discuss NIS integration was referred to the LSD working
group. 

2. Status Update

The draft describing extensions for dynamic entries has completed
WG last call and has been sent to Harald Alvestrand.  

ACTION (Mark Wahl and Tim Howes): send draft on caching, which has completed 
last call, to Harald.

ACTION (Mark Wahl): revise draft on language tags and send to Harald.

Concensus was not reached on the simple paged draft as a standards
track document.  The WG chairs and area directors will discuss with
the authors of that document whether it should be published as 
an informational document.

There were significant comments on the referral draft during its 
last call period.

ACTION (Mark Wahl and Tim Howes): revise the referral draft and 
issue last call again.

However, the area directors will not publish these extension documents
until the recommended authentication methods document has been published.

3. Start TLS and Recommended Authentication Methods

3.1. Preventing man-in-the-middle attacks

An issue was discussed on the mailing list regarding whether client
implementations are required to check that the server to whom they have
established a TLS connection is identified as the subject of the server's
certificate.  This depends on the representation of the server's DNS 
name in the certificate.  It was suggested that there needs to be either
text in the security considerations section, or the client behavior must
be documented.  The client behavior text would depend on the TLS WG, since
it affects all application protocols using TLS.

ACTION (Jeff Hodges): send suggested text to the security area for review,
then submit revised Start TLS draft.

3.2. Format of authorization identifiers

The second issue was whether authorization ids used in the SASL credentials
field must be Distinguished Names or not.  A Distinguished Name is 
syntatically unambiguous, however some simple clients may not know their 
own Distinguished Name and would want to use a more basic username.  The 
SASL RFC requires that the protocol, in this case LDAPv3, define precisely
the form of this ID, however this has not been done in RFC 2251, and so
the issue has arised in the recommended authentication methods document. 

It was noted by Paul Leach that the binding between the authorization id
and the user is not necessarily authenticated.   Also it was noted that 
if a DN is not used, then the mapping of a username to a DN needs to be 
algorithmically represented so that it can be replicated between servers.

A resolution was reached in the third session, following a BOF
between Mark Wahl, Bob Morgan, John Meyers.

ACTION (Mark Wahl): produce new authentication methods draft, then go to
last call.

4. Sorting

Chris Weider described the changes discussed on the mailing list.  These
were to clarify the handling of entries which lack the attributes being
used for sorting, and to note that sorting is not done across servers 
where there are referrals.

ACTION (Chris Weider): reissue draft with these changes, then the WG will go 
to last call. 

5. Access Control Requirements

Ellen Stokes summarized the open issues with the document.  These are 
primarily whether there could be explicit denials, and nested groups.  
The draft currently suggests both should be not-recommended.

5.1. Nested Groups

Richard Huber noted that nesting is useful to prevent a proliferation
of groups.  However, with nesting there can be side effects not known
to the administrator of one group if a group under another administrator's
control is modified. Jeff Hodges suggested it was the responsibility of 
the deployers to ensure that consistency was maintained, rather than the
responsbility of the specification to prevent it from occuring.  

5.2. Explicit Denials

Paul Leach raised the issue from the WEBDAV working group, that it is
hard to disallow access if there are multiple access protocols as there
is no common ACL format.

Richard Huber also suggested that with inherited access control it is 
more necesary to have negative rights. 

Tim Howes and Bob Blakely polled the meeting and determined there was
rough consensus on this topic.  

ACTION (Bob Blakely) revise the document to allow explicit denials.

6. Access Control Model

Deborah Byrne described their proposed access control model, for which
a draft had recently been sent to the list.   Each access control subject 
has security attributes which describe the permitted actions.  Attributes
are grouped into classes by sensitivity.   Inheritance would be used to 
eliminate the need to store access control information with every entry.

7. LDAP Signed Operations
 
Pat Richard listed his planned changes to the LDAP Signed Operations 
document.  This would be to include a reference to PKCS #7 for the 
definition of the LDAP signature, and make the signature structure more
extensible.  There was also inteerest in merging with the draft from 
Vesna Hassler.

ACTION (Pat Richard): produce updated draft.

8. Persistent and Triggered Search

There are still two proposals for client notification of search changes,
however the proposals have slightly different properties and each is 
suited for particular ranges of applications.

Russell Weiser pointed out that the documents require more work in the 
security considerations sections, and Harald Alvestrand reminded the authors 
to be sure to discuss the implications on scaling as the number of users and 
the number of updates increases.

ACTION (Mark Wahl and Mark Smith): revise both documents and reissue.

9. Scrolling

David Boreham described the current state of the virtual list view draft.
The document currently requires that sorting also be used simultaneously,
and there have been some comments on this as the simple paging draft did
not require sorting. 

ACTION (David Boreham): revise draft and reissue.

10. C API

Mark Smith listed the minor issues which still need to be done: changes to
data type definitions, cleanup of the language regarding when information
should be freed, and integrate comments from Mark Wahl relating to the 
functions added in the latest revision.
 
There is a new function for determining the version of the protocol 
supported by the C API, and Mark Wahl suggested that this should also be
extended to support the detection of API capabilities as there is a 
proliferation of extension API documents.

ACTION (Mark Smith): revise draft and reissue.

11. Transactions

A document was recently published on adding transaction operations to LDAPv3.

There are still some open issues, such as time limits on transactions, whether
neested transactions and operations across referrals are supported, and the
interactions with other protocol extensions.

There was no consenus on whether this should be added to the charter.

End of First Session

Second Session

12. Access Control

12.1 Prescriptive ACLs

Helmut Volpers asked about the capability of defining prescriptive ACI for 
covering a whole subtree of entries. Bob Blakely suggested this would be 
done through inheritance, however inheritance is not yet defined for LDAP.

It may be desirable to allow Prescriptive ACIs to cross server naming 
context boundaries, therefore the mechanism would need to be defined in 
LDAP how changes to one server are reflected to subordinate servers.

12.2. Relationship to X.500 

There were questions why the access control format and model defined in
X.500(1993) was not used.  There was interest in investigating a subset of
the X.500 model for use in LDAP.

ACTION (Ellen Stokes): document an analysis of the relationship to X.500(93)
access control

12.3. Nested Groups and Explicit Denial

Since these topics were discussed earlier, there was a meta-discussion on 
whether there should be optional or extensible elements in the access 
control specification.  If so, how are these handled when there is 
replication?

12.4. Subject Expressability 

It is desirable to be able to express in the model certain subjects, such
as "self" or "everyone".

12.5. Default ACL Policy

The proposal had appeared to describe the default access control policy as 
granting read access to everyone, and Ed Reed and others had suggested that
the default should be changed to granting no access.  In their proposal,
the default access control policy for attributes could be changed by the
administrator.

12.6. Multiple ACL Formats

Two related topics were discussed: whether there should be a mandatory to 
implement access control mechanism for LDAP, and more generally, whether
access control in IETF application area protocols should be defined in terms 
of policy on the objects accessed or policy on the protocol.  

ACTION (Tim Howes): ask for input from the security Area Director.

End of Session

Third Session

13. Authorization ID Format

The third session presented a proposed resolution to the representation
of authorization identifiers in SASL credentials.  In the revised document,
the authorization id should be a choice of "dn:" followed by a DN, 
"u:" for user followed by a UTF-8 encoded username, or a zero length string.

There may be more discussion on the list on the related topics of 
replicating mappings for usernames to DNs, and on server proxying.

End of Third Session



Mark Wahl, Directory Product Architect
Innosoft International Inc. / Critical Angle Inc.