Re: [regrep] Dynamic Discovery/Binding/Invocation

["David RR Webber" <david@drrw.info>]

Evan,

You touch on the crux of the matter here.  For example - people
are attempting to drive XSD as that lingua franca - but it is not
equipped for the job - nor was ever intended to be.

The pragmatic approach wins out over time.  There maybe a
better theoretical approach - but if it is not simple enough
for average business users - it will not succeed.

Essentially there are two approaches - #1 - the UBL one - of
enforcing standardized bricks and palettes, and
#2 - the OAGi style approach that blends some component
standardization with extensibility.  Human nature, and the
nature of competitive business edge - gravitates to #2.

In either case - having uniforms means of expressing this -
that CAM, SCM and Registry bring to the table - reduces
the number of deltas that you are grappling with.

Therefore - simply going thru the exercise of utilizing
OASIS specifications like CAM - brings about
convergence and simplifies the overall task.

Bearing in mind that the domain is eBusiness interfacing,
then you expect that techniques and tools that have
been fashioned in the crucible of business use for the
past 25 years - will work and will be applicable.

If the domain is something like satellite telemetry from
the Mars space probe - then do not expect OASIS
specifications to work well there.

In summary - within domains - OAGi to OAGi, etc,
strong alignment - across domains - OAGi - CIDX,
weak alignment.  Just for the record - domain based
integration is a lynchpin of the BCM approach.

Thanks, DW

----- Original Message ----- 
We have been working for a while on similar problems in our
Automated Methods for Integrating Systems [AMIS] project.  There are lots of
hard problems here such as: discovering semantic links, defining semantic
maps,
building non-lossy tranformations between models/schemas.  This is made
worse by the complexity of the content models for even "simple" transactions
in popular ebusiness message suites such as OAGIS BODs or CIDX Chem
eStandards.

Will it ever be possible?  Perhaps, although under very constrained
conditions with better tools and with standard transaction suites built
with a different design philosophy.  IMHO start with extendable semantic
models of business objects and build process and interface definitions from
these.

-Evan