Re: How namespace names might be used

On Sun, Jun 11, 2000 at 10:25:55AM +0100, David Carlisle wrote:
> > Why can't content negotiation handle this? The CONNEG working group    
> > is looking at being able to express just such relationships...
> 
> Moving from the current situation where the namespace name is "just a
> name", an identifying mark, to one in which the expected behaviour
> is that the namespace name be deferenced as a URI would be a massively
> retrograde step in the direction of proprietary data formats and thus
> a move against the core reason for using XML at all.

I have never, ever suggested that always retrieving the resource should
be the expected behavior. In _most_ cases the application processing
the XML is going to have a priori knowledge of the namespace anyway
and would probably _never_ need to retrieve it. Just because
something is a URI doesn't mean you have to resolve it. You can use
it just to disambiguate....

> An example might help, Postscript files that meet certain extra
> criteria start off
> 
> %!PS-Adobe-3.0
> 
> This identifies them as postscript, and is generally a useful thing.
> It also gives credit where credit is due, and acknowledges that Adobe
> specified the format. However anybody can implement a postscript
> compatible engine to process the file. postscript, despite being
> owned as a language by one company, is an essentially open format.
> 
> If we ignore SVG for now, then an xmlisation of this
> would look something like 
> 
> <PS xmlns="http://www.adobe.com/dsc/3.0">
> 
> Under XML namespaces as currently specified this has essentially the
> same status as the DSC comment on the original. It is a unique
> identifying mark, and credits Adobe with specifying the language.
> 
> Under your proposals the expected behaviour would appear to be that
> in order to process the file you had to connect (or have connected in
> the past, to use a cache) to Adobe's server.
> This would be a bad thing.

I never suggested that. And I'm pretty sure no one else has ever suggested
something that stupid. Again, just because something is a URI doesn't
mean you have to resolve it. In this example your application is
a postscript engine (which, turning postscript into an XML namespace 
would be kinda cool). It probably only knows postscript and thus won't
need to resolve that URI. 

Now, lets extend your hypothetical postscript engine just a little. 
Lets assume that we have written an XML namespace that is Postscript.
Now we extend the engine so that it can do the whole myriad of 
XML and XML representations. Now you have a situation where the engine
can handle _other_, previously unknown namespaces. In this case the
engine can attempt to resolve those other namespaces in order to learn
and extend itself to handle unknown stuff.

That feature still doesn't require you to resolve the adobe URI...

> And its not just postcript. Microsoft (to take a random example)
> are quite capable of specifying good and useful languages for all sorts
> activities. If I want to use one of those languages then I have no
> objection to using whatever namespace name the author of the language
> specifies. But in order to use that language, mappings between that
> namespace name and the required tool configurations are going to
> have to be set up. This happens by looking up the namespace name
> in some configuration file (or as briefly discussed this morning) by
> some global lookup. 

Sure. Happens all the time. This process is almost completely orthogonal
to the issue of getting some unknown peice of XML and possibly
wanting to learn about it via the namespace URI...

> It will not happen by content negotiation at
> microsofts server unless they suddenly start advertising linux
> software. Saying that the way to discover facts about a namespace
> is via content negotiation implies that one organisation has
> control over all aspects of the namespace. This is an absolutely
> bad idea.

No one ever said _all_ aspects of a namespace. Just authoritative
aspects of it. Many search engines to day allow you to search 
them for a given URL to see who else references it. No big deal.
Its part of the architecture to do that....

> Tim Bray, Rick Jelliffe and others have made similar arguments on this
> list.
> 
> > the URI identifies the semantic resource not
> > a particular representation of it.
> 
> But the resource identified by the namespace name is usually not the
> namespace so this doesn't really help in this situation. This has to
> be the case as long as every URI (or as now, every URI reference)
> may be used as a namespace name.

Every URI can be used as a namespace name. Its just that some have some
rather funky characteristics. Its all in how the namespace author 
choses to name their namespace. I.e. I can tell my browser that my
home page is the URI "mailto:michaelm@netsol.com". Does that make
sense? Probably not. Does that mean that the "Home Page" field in
my configuration should somehow be restricted to only http URI schemes?
No....

-MM

-- 
--------------------------------------------------------------------------------
Michael Mealling	|      Vote Libertarian!       | www.rwhois.net/michael
Sr. Research Engineer   |   www.ga.lp.org/gwinnett     | ICQ#:         14198821
Network Solutions	|          www.lp.org          |  michaelm@netsol.com

Received on Sunday, 11 June 2000 10:52:46 UTC