This meeting of the DSML group was held on Wednesday, 18 July, 2001 in Austin, Texas. It was chaired by James Tauber. It was hosted by the Directory Interoperability Forum of The Open Group.
The purpose of the meeting was to agree a strategy for development of the DSML 2.0 specification.
Winston Bumpus, chair of the DIF, opened the meeting, and passed the floor to James Tauber, who reviewed the current situation of DSML. There were then presentations from Access360, Microsoft, Verisign and Novell on their respective proposals for DSML 2.0. The issues raised by these proposals were discussed. There was then a brainstorming session in which ideas of various kinds were put forward. Finally, conclusions were drawn, and a work plan was developed to move the activity forward.
Name |
Organization |
Name |
Organization |
|---|---|---|---|
Gary Anderson | Novell | Mark Kosters | Verisign |
Alexis Bor | Directory Works | John McGarvey | IBM |
Winston Bumpus | Novell | Andrew Newton | Verisign |
Peter Farrimond | Jacobs Rimell | Jeff Parham | Microsoft |
Tony Gullotta | Access 360 | Saurabh Shrivastava | Oracle |
Chris Harding | The Open Group | Skip Slone | Lockheed Martin |
Andy Harjanto | Microsoft | Gavenraj Sodhi | Access 360 |
Nigel Hubbard | Data Connection | James Tauber | Bowstreet |
Nigel Jacobs | iPlanet | Shon Vella | Novell |
DSML-AU-01 [John McGarvey, July 31]
Put forward a specific proposal for improving Information Interoperability in DSML.
DSML-AU-02 [Jeff Parham, July 31]
Consider whether he can be editor of DSML 2.0
DSML-AU-03 [James Tauber, July 31]
Propose to the DSML mail list that there should be weekly teleconferences on Wednesdays at 09:00 US West Coast, Noon US East Coast, 17:00 UK time.
DSML-AU-04 [James Tauber, July 31]
Set up the DSMLweekly teleconferences and circulate details to the DSML mail list.
DSML-AU-05 [Gavenraj Sodhi, July 31]
Make the SAML ommunity aware of developments in DSML2.0, and invite them to submit requirements.
James Tauber said that the DSML 1.0 specification was completed at the end of 1999. It just represents information in directories. When it was developed, the question of representing protocol operations also had been discussed, but it was decided not to do this at that time.
The first half of the year 2000 was spent in moving the DSML activity into OASIS. There were then thoughts on what DSML could and should achieve. James believed that it should focus on adding query and modification to DSML 1. Schema considerations are important, but may not be realistic for DSML 2.0. There is no need to try to do everything in 2.0 - it would be possible to develop a DSML 3.0 later.
Tony Gullotta presented the approach proposed by Access360. This approach is the Directory Access Markup Language (DAML), which evolved in the light of the needs for a common information format of their enRole policy-based provisioning product.
DAML is an XML-based language for representing directory services. It can be transported over HTTPS. This approach was chosen in preference to LDAP, because XML is easier to handle than ASN.1, and because HTTPS can generally pass through firewalls whereas LDAP over SSL is more problematic.
DAML is entirely upwards-compatible with DSML 1.0. (Making it upwards-compatible while still meeting Access360's requirements had not been difficult.)
Jeff Parham presented Microsoft's proposal for DSML 2.0. The essence of this proposal was that DSML 2.0 should be a full-fidelity encoding of LDAP operations and results. DSML should encode LDAP as it stands, and not try to improve on it or to solve LDAP problems. This would allow DSML to make rapid progress.
DSML should be transport-independent, but with a request/response paradigm. Possible transport vehicles should include HTTP and SOAP.
There are some issues that need to be resolved with this approach:
Verisign's proposal was presented by Tony Newton. This proposal is for a layered approach, with DSML being one of several possible standards at the top - Specific Directory - layer. This layer should be supported by a Universal Directory layer, which would provide commonality between different kinds of directory; it would be implemented by XDAP. The Universal Directory layer would in turn be supported by the bottom - Transport - layer, which would define bindings to different transport protocols such as SOAP, HTTP, and BEEP.
An advantage of this approach is that it is not bound to the heirarchical (X.500) directory model, but can accommodate other models, such as the IP address structure.
Shon Vella presented Novell's viewpoint. Novell had submitted its DirXML DTD into the DSML debate as a catalyst to get things moving, and were not necessarily committed to it as an absolute basis for DSML 2.0. In fact, it provides more functionality than is needed in DSML at this stage.
Novell felt that encoding LDAP was a good basis for the DSML 2 standard, and that the Microsoft proposal was a good starting point, though some modification might be needed.
There was no argument against the proposition that DSML should be completely transparent to the underlying transport protocol.
In a straw poll, most participants agreed that the goal should be just to represent RFC 2251 operations in XML. But there were some reservations, including Information Interoperability, and Differences between Servers (see below).
John McGarvey said that a measure of interoperability on a small class of objects is desirable. A prime example of such an object is inetorgperson. It was argued that standardization of such objects is appropriate within LDAP rather than in DSML. However, John maintained his position that improving interoperability of information should be a DSML requirement, and agreed to put forward a specific proposal.
It was agreed that DSML 2.0 should support referrals, but not chaining (which is in any case not defined properly in LDAP).
There was disagreement over whether differences between servers should or should not be visible to clients.
The brainstorm session was captured in a presentation that was displayed to the meeting and modified interactively to reflect the discussion as it progressed.
It was agreed that the Microsoft proposal should be taken as the starting point for DSML 2.0.
Further requirements not embodied in that proposal should be noted and kept in mind for future releases of DSML beyond 2.0.
The work will be done within OASIS and following OASIS procedures. Only OASIS members will participate.
(An OASIS technical committee can declare a specification to be finished at any time; the specification then becomes a "committee specification". To become an OASIS standard, a specification has to be put to the full membership. This can happen once every three months; a specification ready at the end of a quarter can be voted on in the following quarter.)It was felt that three months would be a good, agressive timescale to reach committee specification stage for DSML 2.0. However, it was agreed to aim for the final draft by September 30, to allow the specification to be voted on as an OASIS standard in Quarter 4 of 2001.
The date for "feature freeze" should be the end of August.
It was decided to hold weekly teleconferences, in order to progress the work. Wednesdays at 09:00 US West Coast, Noon East Coast, 17:00 UK time was the suggested timeslot, and it was agreed to propose this to the DSML mail list.
Jeff Parham agreed to consider whether he could undertake the editorship of DSML 2.0. The question of editorship was deferred pending this consideration.
Gavenraj Sodhi agreed to act as secretary of the group and maintain the web page.
Gavenraj will also act as liaison to SAML. An action was placed on him to make the SAML community aware of developments in DSML2.0, and invite them to submit requirements.
Winston Bumpus volunteered as vice-chair of the group, and this was agreed. Alexis Bor will also provide backup to the chair as required.