Re: Collect Proposed wordings (Was: Can everyone be happy?)

>My point is that somebody defining a namespace MUST NOT assign different
>semantics to two names that according to the properties of the URI space
>are the same name.

Reminder: It is not at all clear at this time that the person defining the
namespace is in fact defining semantics in any strong sense of the term. It
is certainly _preferable_ if everyone agrees that a particular expanded
name should be used for specific purposes. But there is nothing in today's
architecture which documents that intent, never mind acts on it or enforces
it.

It may be reasonable to expect that {http://usps.gov/mailingAddress}City
will in fact be used to refer to the name of the city to which the
addressee's mail will be delivered. But if someone decides to reuse that
datatype to specify a city for other kinds of deliveries, that's fine. If
they decide to use cities as the names of a particular level of their
network hierarchy, independent of whether the servers are located in those
cities or not, that's a bit weird but also fine.

And if one really insists on writing a schema in which that same expanded
name encodes cities as numbers rather than names, or uses it as the root
element for a description of the city, or uses it as a flag to distinguish
things that are legally incorporated as Cities (versus Townships or
Villages), or to refer to a fictitious city in an interactive novel ("See
the Milky Way on 7 megaergs a day!").... well, maybe some of these are bad
practice, but it's unclear that any of them are really things that someone
MUST NOT do. And there's certainly nothing today which will prevent them
from doing it, or even aid them in deciding whether it's wise.


As things stand today, Namespaces are still syntax. Very useful syntax for
building rich semantics upon now and in the future. But syntax all the
same.





______________________________________
Joe Kesselman  / IBM Research

Received on Thursday, 22 June 2000 19:47:19 UTC