Which came first, the function or the binding?

Heylas,

Sorry to open this can of worms so late in the day, but the question
arises due to the inclusion of the application data feature in part two of
our specification.

WSDL core, as expressed in part one, provides a [description] of an
abstract [web service].  It includes the capability to [extend] the core
functionality along two separate axes: extensions of the abstract
description (of which features&properties are the most formal
representatives of [extensibility elements]), and mapping, or [binding] to
a particular wire format (and even to a particular instance of a service
located at a specific address).

Part three of the specification provides [binding] extensions for the core
functionality set forth in part one.  It also provides, for each [binding]
extension, specific mappings of certain extended functionality found in
part two, specifically message exchange patterns.  Part two of the
specification provides functional extensions to the abstractions defined
in the core spec, in particular message exchange patterns.  As of two (or
one?) week(s) ago, it also provides one particular extensibility element,
the Application Data feature.

The general question: when new [functional extensions] (such as AD) are
defined, where should the corresponding [binding] extensions be described?
 This is a problem of orthogonality.  Assuming that additional, external
[binding] extensions are defined, the authors of the [functional
extensions] cannot be expected to create mappings for them.  Assuming that
additional [functional extensions] are created, the authors of
previously-established [binding] extensions cannot be expected to create
mappings for them.

Does this imply, then, a class of "micro-specifications", which link one
[functional extension] to one [binding]?

Does this imply that it would be good form, in the presentation of the
mappings of [functional extensions] to [bindings] that appear in documents
authored by the working group, that these linkages should be separated
from the connection of [binding] to [core functionality]?  That these
linkages should, generally, be set off in a single location, and that they
should be designed as a template for authors of [functional extensions],
[bindings], and [bindings of functional extensions]?

Amy!
-- 
Amelia A. Lewis
Senior Architect
TIBCO/Extensibility, Inc.
alewis@tibco.com

Received on Tuesday, 20 July 2004 13:12:50 UTC