Lists Home |
Date Index |
> From: Dare Obasanjo [mailto:email@example.com]
> Support for retreiving document metadata from RDDL documents
> on the web seems
> application specific to me.
I can accept that. I think a PI is fine.
> The addition of that PI would solve one of the issues I had
> with RDDL and as
> for the other (no support for URNs and poor support for none
> HTTP URIs), I'd
> go with something similar to Paul T's suggestion of an XML
> metadata repository
> as described at
The extended link approach I've advocated solves the lack of support for
URNs. (Quite apart from that, you can use XML Catalog to resolve a URN to a
URL. Why reinvent the wheel to solve problems for which there are already
simple solutions?) I mean no offense, but this model Paul T is proposing is
sheer bullshit. A type system for resources? Not anyone can create a type --
they have to be "common and validated"? A model based on centrally managed
and controlled repositories? Making derefencing a URI a virtual offsense?
Everything he is proposing flies in the face of the architecture of the web.
Sorry. I'm not buying into that. Not now, not ever.
Simplify, man! Simplify! The goal here should be to make things simple, and
you do that by building upon concepts that are already familiar to most web
developers. RDDL got that right. It builds on hyperlinking and dereferencing
URLs, which every web developer understands. Those that insist a namespace
URI doesn't mean anything, it's just a name, simply don't get it. It
*should* mean something. Otherwise, we are subverting the success that the
web has enjoyed so far because we are making it too abstract and esoteric
for the average web developer.
Every counter proposal I've seen gets this flat out wrong. It feels like
some people want to take a VW Bug and turn it into a Hummer. Have you
noticed that there are far more people drivng VW Bugs on the road then there
are driving Hummers? It's because a Hummer is pretty unweildy and a pain in
the ass to deal with when you just want to drive to the market.
A proposal that insists that "not anyone can create a type" and that types
must be submitted to centrally managed and controlled repositories does not
have a chance in hell of gaining mainstream adoption.