Re: system IDs and URIs

[John Cowan <jcowan@reutershealth.com>]

Lauren Wood wrote:


> To me, the main reason for having generic URIs is to allow mapping 
> of items such as stylesheets from the stylesheet PI, which isn't 
> defined as a system identifier, and potentially other URIs which 
> also aren't defined as system identifiers. I would not like to lose 
> this ability.


Absolutely.  What I wanted to investigate was not eliminating generic
URIs, but subsuming system ids under generic URIs.

In particular, is there any merit to allowing a URI to be mapped
one way when it is a system identifier and another way when it is
being used for something else?  We rejected making such a capability
general, with separate mappings for stylesheets, XML Schemas,
RDF Schemas, etc. etc.  Why should system identifiers get their
own private mapping?

I don't think that 9401 compatibility is a sufficient argument,
because 9401 catalogs don't allow mapping any URIs that aren't
system identifiers.

Norm says it would confuse users to have only one type of URI
processing.  Me, I think it would confuse users to have two different
types.

-- 
There is / one art             || John Cowan <jcowan@reutershealth.com>
no more / no less              || http://www.reutershealth.com
to do / all things             || http://www.ccil.org/~cowan
with art- / lessness           \\ -- Piet Hein