[
Lists Home |
Date Index |
Thread Index
]
[BCc'ed to www-xml-linking-comments@w3.org from XML-DEV reply]
On Thu, Nov 14, 2002 at 11:03:31AM -0500, Simon St.Laurent wrote:
> IP and DNS are two spectacularly expensive label spaces. I don't think
> that they are the only models worth considering, however. Given the
> likely scope of XPointer, I don't think the wide-open expanses of URI
> possibilities make much sense, and something much smaller and perhaps
> even managed is worth consideration.
Especially in light of interoperability. The goal is still to be able
to point *acurately* and *reliably* to a fragment of an XML resource.
There is no way to go ATM from an {URI}scheme pair to a definition
of the semantic of that scheme. While it's fine from a XML namespace
point of view (where an awful lot of processing can be done on an XML
instance without understanding the semantic of the tags for name and
attributes), this is clearly not the case for an XPointer processor
which must understand the semantic of the scheme to make any use of
it (except skipping it and trying the next scheme if available).
So there is a need for a registry *anyway* (until someone find
a way to express the semantic of the scheme and anchor it to the
namespace name which could take years and years to standardize if
feasible !).
Namespacing the schemes only solves the problem of name unicity,
but it does not solve the need for a registry, at least the proposal
currently on the table does not.
Daniel
--
Daniel Veillard | Red Hat Network https://rhn.redhat.com/
veillard@redhat.com | libxml GNOME XML XSLT toolkit http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/
|