Lists Home |
Date Index |
Rick Jelliffe wrote:
> So the primary utility of a document type should be during
> creation and maintenance phases, especially when starting
> a document type from scratch. During this phase, the issue then
> becomes how can we provide enough resources so that desktop tools
> can be configured? They need to get the house-style versions
> of schemas and stylesheets, templates documents which
> provide typical cases, and vendor-specific code, to allow
> vendors, philanthropists and experimenters.
> I think only DZIP goes anywhere near addressing these considerations,
> so far, supporting RDDL, CATALOGs, vendor-specific code, templates,
> We need to move beyond document types to distributable, extensible!, identifiable
> (and, sure, web-locatable), system-integrator-friendly "XML Applications".
I agree -- somewhat. I think the Java world serves as a good analogy.
You can drop a JAR on the classpath, and all the contained classes are
immediately available to the application. You can also load classes from
a URL (though ideally this should only be used for initial discovery --
once found the resources should be downloaded and accessed on demand
from the local filesystem). XARs and RDDL are complementary, here.
I think, though, this is a use case that points up a weakness of RDDL --
and XARs without something like RDDL are even weaker. Taking RDDL and
just stepping up to extended links strikes me as just what the doctor
ordered for XARs. Using such a format, a XAR can contain resources and
associate them with an arbitrary number of namespaces, doctypes, or
whatever. What I'd like to see: a single file to look for (not check for
index.html, then index.htm, then default.html, then default.htm, etc.)
-- similar to a JAR manifest -- that helps me locate all of the other
resources in the XAR and tells me their nature and purpose (and what --
if any -- URIs they are associate with). I don't see how XAR achieves
this objective without using extended links or something equivalent.
And regarding that "something equivalent" comment, I would say that
XLink syntax is the right way to go -- not RDF. Ultimately, the format
should have some faint hope of getting mainstream adoption, and that is
best achieved by building in an iterative, evolutionary fashion upon
concepts with which the average web deveoper is already familiar. Every
web developer understands hyperlinking at some level. Almost every web
developer has completely ignored RDF, and using RDF pretty much kills
hopes of mainstream adoption.
Why not have an index.xml or index.xhtml in each XAR that adopts RDDL
conventions, but using extended rather than simple links? The same
syntax for the extended link could also be used in linkbases, enabling
applications to permit the user to associate additional resources with a
namespace or doctype without having to package a XAR or put RDDL on a