Re: Object Role Modelling (ORM) or UML or ?? for designing Schemas

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

Danny Ayers wrote:

> This seems to me to be suggesting that the markup itself has no semantic
> characteristics, in other words doesn't contain any information itself - I
> may be way off the mark, but I would have thought that the semantics of the
> markup would be anything but local to each node. Surely there will be some
> semantics associated with that markup, whether it is implicit in the item
> (which I have trouble with) or is provided by means of the citation of other
> resources that define the pieces of data and their relationship(s). In other
> words the meaning of an item will be 'universal', though communicated by the
> markup, and only the processing or interpretation of the item will be local
> to the internetwork node. Of course this may be different at different
> nodes - and it is likely to be the case that each node only uses a subset of
> the information conveyed by the item. The node can't elaborate anything new
> out of the item without interpreting it with reference to other
> information - and that will usually be greater than, and always different
> than, the semantics carried by the syntax itself.

Thank you for your kind appraisal. It seems to me that in the course of this
final paragraph you move visibly closer to my opinion. There is no question that
the creator or transmitter of markup syntax may intend specific semantics. The
user or recipient of that markup, however, may not understand that intent, or
may disregard it, precisely because his use for the data transmitted, or his
concept of the structure into which it fits, is different. As you note, the
recipient might use only a subset of the data transmitted, which substantially
changes the intended information conveyed. The recipient might equally well
require for his own unique purposes a superset of that intended information. The
ability to construct that superset implies that the recipient comprehends the
syntax received within a larger understanding (a taxonomy or an ontology, if you
wish) than that intended by the transmitter. The point is that, whatever his
intent, the transmitter of markup syntax across an internetwork should not
expect that syntax to be understood or processed in a given way.

It may be useful at this point to remember that we are not talking here about
semantics, syntax, or any relationship between them in an abstract manner or a
universal context. Our subject is the specific business of marking up text which
will be processed by software. The semantics which will be understood by the
recipient or user of that text are those which are elaborated from the markup
syntax by a process which operates in the potentially unique environment of that
recipient. That process may be utterly idiosyncratic, or in any case may be
entirely different from the intent or expectation of the creator or transmitter
of the markup syntax. The point is that the semantics understood by each user of
the markup syntax are determined by the process which elaborates those from that
markup but also from other sources which may be understood only by that process
itself. In short, it is pointless at the time of marking up the text to
determine its semantics, in the sense of some holistic understanding of that
text at the time of its use. Yet it is equally fruitless to attempt to craft
markup devoid of semantics, or which does not reflect some intent of its
creator. Even such mundane matters as the boundaries of markup elements imply a
particular articulation of sememes, and both the hierarchical structure and
document order of markup impute significant meaning to the relationship of
semantic as well as textual components. There is little point in trying to hide
the intent of the markup creator, but that creator should not attempt to impose
that intent upon the various users of that markup syntax.