Re: [Ecrit] Profiles (was: Consensus Call) - namespaces

[Henning Schulzrinne <hgs at cs.columbia.edu>]

Maybe we need to restart this discussion, as I'm lost. I still  
haven't heard a single technical argument as to what additional  
information a profile name, outside the XML, contains that would not  
be visible in the namespace and tag declarations. I was taught that  
XML documents could stand alone, without additional meta information,  
once you have pointers to the namespaces. Do you have a citation that  
indicates that this is wrong?

I'm looking at Walmsley, "Definitive XML Schema" and it says:

"Instances can be related to schemas in a variety of ways ...

* ...
* dereferencing the namespace. The namespace can be dereferenced to  
retrieve a schema document or resource directory."

In this case, this is a URN and all relevant schema will be known to  
the receiver and we have a one-to-one relationship between namespace  
and schema, so there's no ambiguity. (Walmsley points out various  
reasons why an HTTP-like namespace identifier may not be a good idea,  
but none of them seem to apply here.)

Also, as I mentioned before, XMPP seems to follow roughly the same  
model as 4119. E.g., the core spec defines a schema and namespace  
urn:ietf:params:xml:ns:xmpp-sasl and urn:ietf:params:xml:ns:xmpp-tls.  
There is no profile indication, as far as I can tell, even for the  
various extensions (say, namespace jabber:x:data).

Can you please provide a reference or other authority that supports  
or explains your view?

Henning

On Apr 26, 2007, at 11:03 PM, Andrew Newton wrote:

> On Apr 26, 2007, at 10:53 PM, Henning Schulzrinne wrote:
>
> > I've asked a few times, but why would LoST care that the civicAddr  
> > is derived from DHCP as opposed to, say, entered manually?  
> > Obviously, the current profile doesn't convey that information,  
> > either.
>
> I don't think it should care.  The point wasn't that it came from  
> DHCP, the point was that it was a 3825-like compound location.   
> Sorry if that was confusing.
>
> -andy


_______________________________________________
Ecrit mailing list
Ecrit at ietf.org
https://www1.ietf.org/mailman/listinfo/ecrit