[
Lists Home |
Date Index |
Thread Index
]
Your <?XAR docbook?> PI is pretty much the document type specification I'm
thinking about in my article. The difference is that I would not force the
resource directory and the archiving mechanism to be bound. For me (but I'm
just thinking about it, so I lack plenty of justifications), a document type
name should be an URL, that points to a resource directory. This resource
directory would allow me to fetch schemas, stylesheets, human-readable
documentation, Java archives (i.e. plugins for browsers in the form a Java
Applets), a XAR archive, and, etc.
We could even have RDDL documents for namespaces, why not ? This resource
directory would also serve as an URI resolution mechanism, thus enabling the
resolution of non-URL namespaces name, so that even those namespace could
benefit from RDDL. But, of course, RDDL would have a warning at the
beginning of the spec : "WARNING : do not use RDDL to obtain schemas for
your documents. Use the resource-directory associated to the document type
instead. A tough battle has been fought on those grounds, so please don't
disturb the ghosts".
I would not try to put too much in the same specification, and try to
separate adressing and URI resolution concerns from packaging concerns. But,
as I've said before, I'm only thinking about it for now.
It's amazing to see how we can agree on this and have the same ideas, yet
having had to fight so fiercely on the list about RDDL. I hope the
discussion with progressively cool down, so that we'll be able to enter the
engineering aspects :).
Cheers,
Nicolas
----- Original Message -----
From: "Rick Jelliffe" <ricko@allette.com.au>
To: <xml-dev@lists.xml.org>
Sent: Sunday, January 20, 2002 10:18 AM
Subject: Re: [xml-dev] Re: Flexible Schemas (was RE: [xml-dev] The task to
be solved by RDDL)
From: "Dare Obasanjo" <kpako@yahoo.com>
> I believe the thrust of Nicolas' question is where to find the RDDL
document
> for an XML file that utilizes names from multiple namespaces. No one has
> proposed a *concrete* answer to that question yet although the designers
of
> RDDL may consider this to be beyond the design goals of RDDL which is
> reasonable.
RDDL answers "what is at the end of the namespace URI?" The resource
at the end of the namespace URI for a RDDL document is an XHTML
document. So the RDDL docuemnt that the XHTML namespace URL
locates will not help us with RDDL. Instead, it gives us the typical
or general documentation.
And the schemas that may be located through a RDDL directory may
also be the most general, or most typical, ones for a namespace rather
than the particular ones we are interested in.
I don't see it a fault of RDDL in this. It has been pretty well accepted
since the big discussions of HTML's 3 DTDs that a namespace
does not mean a schema. This has two connotations: first that
the URL is too valuable to tie it to a single terminating type of
resource only (e.g. the schema); but second that there can be
multiple schemas possible (even in the same schema
language) for the same resource--a namespace is not a language
but a language family.
Since a namespace is not a particular
schema in a particular schema language we need to both
try to support as much plurality as we can. So RDDL provides
links so we can locate different language versions of a
schema which (in some sense) is regarded as typical or canonical,
while well-designed schema languages (such as W3C XML Schemas
and ISO DTDs) support mechanisms for associating variant and
particular schemas to a namespace (e.g. schemaLocation in XML
Schemas). Each covers one aspect.
So there is no mechanism for finding RDDL variants from a namespace
URI yet. But that is not a problem with RDDL, it is just something
extra that is needed.
In DZIP[1] aka XAR, I am trying to resolve a very similar problem.
How to bundle and transmit the particular application resources
for a document type (which could have a characteristic namespace).
There is as yet no method of tieing the XAR to a URL. Instead
we just have a simple PI in the document
<?XAR docbook?>
which says to an application "use the application with the nickname
docbook, as resolved by the application itself from whatever repositories
or built-ins". I.e. track down a docbook XAR (perhaps from your
system integrator), then load it.
Incedentally, it is only by having some mechanism like this, I believe, that
we can move beyond a passive acceptance of Microsoft's Derek Denny-Brown
comment on this list that "Lock-in is inevitable, no matter what you do. "
...
"The key to XML is not it's ability to avoid lock-in.
It is it's ability to facilitate interop, and loosely coupled systems. "
Why should we just lie back and enjoy it?
With XML, if we reduce ourselves to just using the namespace URI
directly for tracking down resources, we can only really get
fairly vanilla and generic resources. Generic resources from namespace
URIs are no threat to the kind of lock-in that Denny-Brown espouses.
So I think Paul T and Nicholas are actually asking some good questions,
even though they may be too ferocious that apples are not oranges.
A namespace is not a document type, and we need better support of
document types in order to build shrink-wrapped XML desktop applications.
This is exactly where XAR can help. Integrators can bundle all the RDDL
resources for a particular document type/namespace, and as well a system
integrator can distribute appropriate application configuration files and
binaries
with it. We are prototyping this at the moment in our Markup Editor at
Topologi: one can browse servers for various applications, download them,
and they load the correct DTDs, schemas, RDDL documentation, and
everything else.
I hope other vendors will support XAR too.
Cheers
Rick Jelliffe
Topologi, Pty. Ltd.
[1] http://www.topologi.com/public/dzip.html
-----------------------------------------------------------------
The xml-dev list is sponsored by XML.org <http://www.xml.org>, an
initiative of OASIS <http://www.oasis-open.org>
The list archives are at http://lists.xml.org/archives/xml-dev/
To subscribe or unsubscribe from this list use the subscription
manager: <http://lists.xml.org/ob/adm.pl>
|
