[provision] Monday's Meeting: Lets talk about the models

[Darran Rolls <Darran.Rolls@waveset.com>]

Folks

As you all know, tomorrows PSTC meeting is dedicated to a motion and
resolution on Gerry's proposed changes for SPML.  Over the last two
weeks we have seen a very lively and I must say interesting debate on
this subject.  As chair, I would like to thank both Gerry and Jeff for
making the considerable effort required to mature this discussion in a
very short space of time.  

In matters so technical, two weeks *is* a very short window.  Obviously
neither Gerry's proposal, nor Jeff's alternative/compensatory changes to
the current draft, has been conclusively thrashed out by the committee
at large.  Accelerating the debate has been a necessity brought about by
our current delivery plans for the 1.0 specification.  

This debate and exchange of implementation models should and will
continue in a form to be decided by the committee on tomorrows call.
What lies before us is a simple decision.  Does the alternative model
proposed by Gerry call for the postponement of our current direction and
1.0 specification draft process?  I shall pose this very question in a
motion from the chair when to open tomorrow's official committee
meeting.

As background for this motion, I wanted to present a personal abstract
of the differences between the two models in an attempt to ensure we are
all on the same page and addressing the right issues.  I do not claim
this abstract to be in any way definitive (or accurate ;-) outside of
being my personal attempt to summarize the essence of this debate.  I
present it here for committee review and as guide/background for
tomorrow's discussion.


Where is Provisioning in the Current Model?
-------------------------------------------
As many of you are no doubt aware, early in this effort, we took the
decision to base SPML on a very simplistic core operations model.  In
this model the semantics of an individual provisioning action lay in the
definition of the underlying service schema.  We provided a small number
of generic operations (Add, Modify, Delete, Search) and an open model
for the definition and discovery of that schema as a set of simple
name=value pairs.  To compliment this, we added ExtendedRequest to allow
individual providers to define new operations that did not overlap with
the core operations.

Quite obviously (and openly) this model is schema centric.  Much like
DSML (hence LDAP), SNMP and other like protocols, our intention was to
place the object models of the participating systems in their definition
and support of underlying service schema.  Yes we added a rich
request/response model, support for asynchronous requests and a number
of other features our members felt essential to make this model work in
our problem domain.  Beyond that, we accepted the basic fact that with
our chosen model, our next big task was defining standard service schema
as the definition of the provisioning domain.

Where is Provisioning in Gerry's Model?
---------------------------------------
Gerry's proposal has been much more directly focused on provisioning
actions.  Rather than restricting SPML to a generic set of
non-provisioning centric operations, the proposal presents a more
flexible model in which SPML would consist of a defined set of basic
CRUD operation (like the current model) but would also define additions
provisioning centric operations like say "SuspendAccount".  The idea
being that the provisioning problem domain is controlled and well
defined so why not model that domain in more specific operations and a
more open schema definition model.  To support this Gerry proposed an
extensible operations definition model based on WSDL and a service
schema implementation that made use of XML Schema to provide support for
complex element definitions in underlying service schema.  

Obviously we have not worked through the complete implementation for
this new model; time has simply not allowed this process to complete.
Gerry provided a number of very complicit worked examples and we have
seen much debate on the specifics of the examples.  Beyond these
implementation specifics, I believe Gerry calls for the object model for
service provisioning to be more explicitly defined in its core
operations, and with that for the support of complex native XMLSchema
"types" as elements in the underlying target provisioning schemas.

What are we deciding?
--------------------
Quite simply, in light of Gerry's proposal, does the current model and
draft specification have the support of the committee?   If it does,
then our challenge is to learn from this exercise, take issues, note
objections and move on to complete the 1.0 specification.  If it does
not then we must very quickly access the impact this has on our current
deliverables and dependencies and rapidly devise a plan to execute on
these conclusions to the satisfaction of all concerned.

Next Steps
----------
Come to the meeting tomorrow 3/17.  We will review voting status for all
members and hold an open vote on this issue.

See you there.

=========================================================
Darran Rolls                      http://www.waveset.com
Waveset Technologies Inc          drolls@waveset.com 
512) 657 8360                     
=========================================================



----------------------------------------------------------------
To subscribe or unsubscribe from this elist use the subscription
manager: <http://lists.oasis-open.org/ob/adm.pl>