Information Technology Reference
In-Depth Information
where to draw the boundaries?" \Well, that's one of the value judgements that
an architect has to make, isn't it? What do we want to treat as a platform?
Whatever we choose, below it multiple technical solutions can be selected to
suit different situations. Over time there is a gradual upward movement, as
we understand how to standardize and integrate more complex technologies."
She held her cup out for a refill.
\All right then," he said, picking up an oatmeal biscuit, \what about the
boundary between the information and computational models?" \Well, we
know it's normal to import chunks of the information specication and ex-
tend them in the computational view, but computational objects are generally
coarser grained, so there are information objects that become just properties
in the computational view. Actually, I should have raised this just now when
you were talking about the SIM object as a parameter; it's not actually a
computational object." He frowned and looked straight at her over his cup.
\But why? What's the basis for making the distinction?
Claire thought for a moment, then leaned back, running her fingers through
her hair. \What's the smallest object in your models?" \Well, I guess it would
be a specialization of one of the basic types; let's say an integer specialized
as an employee's age." \OK, and would you expect to migrate the age object
to a dierent system and interrogate it remotely?" \No, that would be much
too inecient." Claire smiled. \Well, that's the basis of the distinction. A
computational object is a candidate for distribution. That doesn't mean it
will actually be moved about on it's own, but that we would be prepared to
consider doing so at the design stage." \That seems a bit vague to me," he
said. \Yes, it's another architectural judgement call, and depends on what's
reasonable use of the technology available, but it's generally pretty clear in
practice. We tend to be a bit generous in making the decisions, because
there are ways of constraining groups of objects so that they are managed
together when the actual distribution decisions are taken in the engineering
and technology models."
\Don't you sometimes feel a bit frustrated not to be making hard and fast
decisions about that sort of thing?" \Goodness, no," she said, \I'm aiming for
stable designs that don't have to be revised every time there is a new wave of
technology to cope with. The applications are too complex to keep reinventing
them all the time." She sat up and looked back to the screen. \What about
the set of identifiers on the SIM. Do we need a common model of the roaming
control data?"
4.1
Designing with Computational Objects
The goal of the computational viewpoint is to model the basic functional-
ity of the application, the services it offers, and how these services are realized
