Re: "Uh, what do I need this for" (was RE: XML.COM: How I Learne d toLove daBomb)

["W. E. Perry" <wperry@fiduciary.com>]

Michael Brennan wrote:

> Developers still need to write business logic, and they need to get the
> information from the XML document into data structures better suited to
> supporting that business logic. All of the APIs in the XML world are unsuited
> for this. They force the developer to mold their logic to fit an API that is
> only suited for modeling a document structure, rather than letting the
> developer mold their data structures and APIs to align with business concepts
> and processes.

Absolutely! In short, they need to instantiate the specific concrete data which
their own processes require. As you point out, they do not need to (and
rationally should not) be creating a document object abstraction (or IMHO a
PSVI) which is not directly processable precisely because that does not provide
the physical instantiation of specific data which their business software
demands.

However, once that specific data structure is instantiated, it will almost
certainly be different from the corresponding data instantiated by their
transactional counterparty at the other end of the XML message. Existing
transaction processing protocols are predicated on a data structure shared by
the parties (and on the underlying assumption that the two parties are doing
congruent things in the transaction). In the case you describe, any general
transaction protocol will have to operate with the very different assumption
that each party is not only handling data in its own way for its own purposes,
but that the concrete data which each regards as the substance of the
transaction is materially different from its counterparty's data. That
recognition undermines the 'agency' assumption--that the two parties to a
transaction process are carrying out a brokerage, essentially of a common data
structure whose fungibility derives from its pre-agreed form. Where the specific
data structures which the two parties instantiate at the moment of transaction
are different, a 'principal' rather than 'agency' model of the the transaction
is required and must form the basis of any general transaction protocol which
might be usable in those circumstances.

Respectfully,

Walter Perry