RE: [ebxml-msg] Summary of where we are with the conformance clause (CC) for multihop

["Jacques R. Durand" <JDurand@us.fujitsu.com>]

The position in the TAB about conformance clause semantics is (now that a conformance clause is mandatory):

 

- "claiming conformance to the specification" is something that can only rely on the conformance clause in the specification document.

 

- this conformance clause may define levels/profiles, so a developer can claim conformance to the specification "according to level L or profile P" or the like.

 

- Any group of users can later on define "profiles" of a specification, in a separate doc. These profiles - if in OASIS - define their own conformance clause(s). Implementers and/or users of such profile can then only "claim conformance to the profile X" of the specification S. Not directly to the specification S.

 

That confirms the need to define a "minimal" clause.

For maximum interoperability and simplicity, I'd favor option (a) below: If some products or users want to only support pulling the intermediary on the last hop, they could define a separate "pulling" profile and claim comformance to this profile of multihop specification.

- If however we have good reasons to believe that (1) hub-and-spoke will be a quite common topology, (2) pulling will be such a common practice that intermediaries should not be required to support "push" in order to conform to Part 2, then we should consider (c).

 

-jacques

---

From: Jacques R. Durand [mailto:JDurand@us.fujitsu.com]
Sent: Thursday, March 04, 2010 4:44 PM
To: ebxml-msg@lists.oasis-open.org
Subject: [ebxml-msg] Summary of where we are with the conformance clause (CC) for multihop

 

Summary of where we are with the conformance clause (CC) for multihop:

 

1- informal consensus is that the conformance clause can address a "minimal profile" which is easy to implement, while reflecting most commonly expected use.

 

2- current clause for "Simple Multihop" sets the bar too high w/r to this goal:

Supporting both MEP bridging models (described in 2.5.3) by Intermediaries: "push-on-push" and "pull-on-push" should not be required.

Instead, only one should be required.

 

3- Three options:

(a) the CC only requires push-on-push to be supported. (and requireing endpoints to support MEPs that push only)

(b) the CC requires either "push-on-push" or "pull-on-push" or both to be supported.

(c) two CCs are drafted (meaning two ways to conform) that are very close to each other:

- one requiring "push-on-push" for intermediaries (and requireing endpoints to support MEPs that push only)

- the other "pull-on-push" (and requireing endpoints to support MEPs that push for sending, and pull for receiving)

 

4- Problem with (a): it forces Intermediaries designed only for pull-forwarding, to also support push-forwarding.

- Advantage of (a): it is a simple clause, that supports interoperability.

- Problem with (b): not helping interoperability... options inside the Clause give little meaning to claiming conformance.

- Problem with (c): multiplies the conformance clauses (or conformance profiles), with little differences. Ideally these variants should be controlled by product configuration, not by product feature implementation choice.

 

 

Jacques