RE: [dita] strategy for identifying DTD/schema artifacts

["Grosso, Paul" <pgrosso@ptc.com>]

> -----Original Message-----
> From: Park Seth-R01164 [mailto:seth.park@freescale.com]
> Sent: Thursday, 2009 September 24 11:09
> To: dita
> Subject: [dita] strategy for identifying DTD/schema artifacts
> 
> I'm in the process of cleaning up the SIDSC specializations and
> catalog.
> 
> My question is, what is our going-forward strategy for identifying
DITA
> DTD/schema resources?
> 
> Should I use a public ID such as "-//OASIS//DTD SIDSC Component//EN"?
> Or should I use oasis namespace identifier in a URN, such as
> "urn:oasis:names:sc:sidsc:dtd:component.dtd:1.2"?

For DTDs, use Formal Public Identifiers (FPIs) such as 
"-//OASIS//DTD SIDSC Component//EN".  

For XSDs, use URNs such as
"urn:oasis:names:tc:dita:spec:learning:xsd:learningContent.xsd"

> 
> What are the guidelines for constructing public IDs?

I would follow the example of the Learning and Machinery
cases for both DTDs and XSDs as can be seen in the OASIS
distributed catalogs.

> 
> The catalog shipped with the OT

[I assume that's just the catalog shipped as part of the DITA 1.2
spec.  In any case, use the DITA 1.2 distribution as the authority,
not the OT.]

> has schema URNs as a system ID; can a URN be used as a public ID?

Technically yes, as long as the URN doesn't contain characters
disallowed
in a public id--see http://www.w3.org/TR/REC-xml/#NT-PubidChar .
But it wouldn't be a Formal Public Identifier, and that's usually
what one wants to use for identifying such public entities, and
then the catalog would have to use a PUBLIC entry instead of a
URI entry, and the would just be confusing to most people.

So I'd recommend that you use URNs.

> 
> Should every MOD and ENT file also have a publicID/URN?

Yes.  This was an issue raised by Eliot quite a while ago, and
we agreed that you need to have an FPI/URN--and appropriate
catalog entry--for every file.

> Or is it
> sufficient to provide a publicID/URN for the top-level shells and use
> relative URIs as system IDs to reference MOD/ENT files?
> 
> I suppose the URN would only be available for organizations with a
> registered namespace identifier, so most user local shells and
> specializations would not be able to use URNs. Aesthetically, a mixed
> approach for public identifiers would be unappealing, but harmless, I
> presume.

One doesn't have to have a registered namespace identifier to 
use URNs.  Regardless, for local use, one can use just about 
anything that works.  Use of relative system identifiers is 
common, and a mix of URNs and public identifiers is okay.  
In fact, with appropriate use of a catalog, just about any 
identifier is okay in a local setting.

paul