[
Lists Home |
Date Index |
Thread Index
]
Paul,
I've just spent some time working back through your
posts, and the previous discussion to attempt to understand
where the disagreements are.
It seems that the fundamental point of disagreement is that
you're asking "what is RDDL for?, what problems does it solve?,
what applications does it let me build more easily?". And in
particular you're considering it's applicability to distributed applications
where some negotiation is required (cf: Lens original thread).
You've asserted that the origins of RDDL are political and not
technical in nature and that any suggestions as to how RDDL might be
used in that context are purely scientific. In effect that RDDL is
a solution to a political problem ("Whats at the end of a Namespace?"),
and is being used to propose technical solutions in other scenarios.
Here I disagree. RDDL is a solution to a problem that you helped identify,
and which Mike Champion dubbed the 'Tool X Horror Scenario' [1]. In your
opinion the loophole that would allow NS URIs to be abused should be closed.
You also argued that it would take 'years' to identify what should be
put there if anything. Others believed letting it point to a directory of resources
would meet everyones needs. I don't see any politics involved, just a
different opinion on the best way to avoid the Horror Scenario.
We can reasonably disagree over whether you see this as a political
issue. I see RDDL (or an alternative) as more of an interoperability spec, that
would avoid the creation of ad hoc solutions. As you disagree that this
is a useful problem to solve, it's not surprising you don't see any
value in RDDL.
You've also stated that some of us, myself included, have been misleading
developers by telling them that there's nothing at the end of a NS URI, but
then turning round and telling them to use RDDL. I still believe the former
to be the letter of the spec, but believe the latter is required for the
above reasons. Any claim of 'spy vs spy' is a little insulting.
You've agreed that something *like* RDDL (a way of discovering
resources) *is* required. But have raised issues about caching, certification,
etc. There's no disputing that. But this seems to have more
to do with the resources themselves ("How do I know this Java class
isn't going to damage my system?") than the directory format. Note that
it's also previously been pointed out that RDDL docs could be cached
locally if one deferences the URI via a Catalog.
As far as concrete feedback on RDDL goes, I see you've made the
following points:
- rddl:resource and anchors have redundancy, you'd prefer something
like annotations on the anchor directly. Your rddl-hook attribute.
- that defining something called 'purpose' with an attribute called 'arcrole'
is confusing.
I'm not sure I really do understand your other claims about how RDDL
interacts with Namespaces. I have a feeling that your saying that a
single general purpose schema for a namespace isn't useful for
circumstances where those elements may be mixed freely with others
from other Namespaces. If so I agree, but I don't see what RDDL
has to do with this.
To me it suggests that elements intended to be used in this way should
not used closed schemas. This is one of the reasons I like Schematron so
much.
Cheers,
L.
[1]. http://www.xml.com/pub/a/2001/01/10/rddl.html
[2]. http://lists.xml.org/archives/xml-dev/200012/msg00662.html
|
