Re: Link Indirection
W. Eliot Kimber wrote:
> It would be simple enough to say something like "lacking a DTD, XML
> processors shall associate the first specification of an -xml-arctype
> attribute encountered in a document to all elements of the same type that
> do not exhibit an -xml-arctype attribute." (Where "-xml-archtype" is
> whatever attribute we choose for naming XML-defined architectural element
> That seems like a pretty intuitive behavior to me.
Wouldn't it be intuitive to say, "if there is a doctype and I have
seen this type before, I know what the links are?" IOW, a policy based
on known shared declarations? The author/organization has options for
this. They can:
1. Send the DTD/Request the DTD
2. Put the type information in the instance.
3. Read the Doctype and trust. No doctype, no attributes,
use local default.
4. Name the stylesheet in the XML instance.
5. Put the stylesheet in the instance (yeachhh!)
6. Put the DTD and the stylesheet in the instance (double yuk).
7. Send a multipart type with parts requested.
The bigger the domain of agreement, the more likely the agreement
requires enforcement. HTML is that. Agree on this set; process this
set. SGML/XML widens that. It says, find your agreements about
this set and apply them. Thus, we need DTDs and stylesheets.
Sometimes we need one or the other or both; sometimes neither.
How much leeway do you want? Could we use a progressive laissez-faire
approach that makes the web server manager the controller of the
complexity of XML transaction types they want to manage?
What DOES the XML parser tell any other engine about XML hyperlinks?