[
Lists Home |
Date Index |
Thread Index
]
W. E. Perry scripsit:
> Which of those [postal codes, viz. on the letter or the envelope]
> is useful (or how either of both may need to be combined with
> other data) for the particular processing which the recipient may need
> to perform is inherently unknown to the sender. As a specific example,
> whether or how the recipient might use the post code embedded in the
> document to disambiguate the semantics of some noun appearing in that
> document from the semantics of the same noun coming from an apparently
> different source cannot be known to the sender except as the effect of
> a priori agreement which that sender is confident that the recipient
> adheres to.
I agree with all of this, but I suggest that what the sender's confidence
about the recipient is far less interesting than the recipient's confidence
about the sender. Senders should 1) agree to act in publicly specified
ways, and 2) act in the way they have agreed to.
The basic model of RELAX NG validation, e.g., works like this. I send you
a RELAX NG schema which specifies the structure of the documents I am
going to send you. You then may, if you see fit, validate those documents
against that schema and reject/complain/except on those that don't validate.
Alternatively, you may provide the schema and require me to conform to it,
but I agree that this case is less scalable unless I am a monopsonist.
> Oh, how I should like to believe this. Unfortunately it unravels
> rapidly in the real world--your example below being only the tip of
> an iceberg. In the same minds which want to disambiguate via static
> 'namespaces' the semantics of a noun used in one context from those
> of the same noun used in another, it seems to be a hard-wired assumption
> that *particular* semantics can be conveyed by the use of a given noun
> with a specific namespace.
Not in my mind, at least. Abusus, I say again, non tollit usum.
But what is the usus? See below.
> On the other hand, if we start with the converse assumption--that
> semantics are not conveyed in the transmission of a document from
> sender to recipient--we will quickly lose interest in namespacing the
> nouns within document content, precisely because we understand that
> the point of view of the sender thus expressed is is not binding upon
> the recipient and for the purposes of the recipient's processing of
> the document is likely to be contradicted by the specific provenance
> revealed in obtaining that document through the internetwork.
I think that this is quite true if the sender is monolithic. However,
if the sender's output contains structured content obtained from elsewhere,
namespaces provide a convenient facility for separating the layers. For
example, an aggregator may use NewsML 1.1, which has a namespace unique
to itself, to package almost any other kind of XML-expressible content.
By using a distinct namespace, there is no chance that the container and
the contents will be confused, it being illegal to wrap NewsML with NewsML.
The same thing is true for RELAX NG or W3C XML Schema with respect to
annotations, and for RDF as well.
--
He made the Legislature meet at one-horse John Cowan
tank-towns out in the alfalfa belt, so that jcowan@reutershealth.com
hardly nobody could get there and most of http://www.reutershealth.com
the leaders would stay home and let him go http://www.ccil.org/~cowan
to work and do things as he pleased. --Mencken, _Declaration of Independence_
|
