[
Lists Home |
Date Index |
Thread Index
]
Quite awhile back I wondered out loud if it was possible to merge
declarative schemas and object-oriented languages:
http://www.xml.com/pub/a/2000/05/03/lowery/index.html
To my mind, the biggest problem with OO languages is that they convolute the
data model with the processing model (BTW, this can be done in XML just as
well: "What is JDF?" http://www.cip4.org), which makes it hard to
deconstruct one from the other.
If, on the other hand, one separates out the internal data structures and
their declarative constraints from that of the processing model, you have
several advantages:
1) constraints on the internal data model are clearly articulated
2) the data model is shareable (schema as contract)
3) methods can be loosely coupled to data structures, yet still retain OO
characteristics
- methods are bound to the data structures they access/mutate
- more than one set of methods can be used on the same data
structures.
- maxim: data models stabilize, process models continually morph
4) simplifies transactions, rollbacks, and change notifications based on
data model manipulations
I think the efforts at XML data binding pretty much follow this path, but
the linkage between any arbitrary OO language and the physical data model of
XML is fraught with inconsistencies and lack of granularity control. In
order for this type of thing to work well, I think both schemas and methods
have to be brought together into one language.
Two cents,
Jeff
|
