RE: [dita] DITA Proposed Feature: Extensibility of DITA through new attributes

["Paul Prescod" <paul.prescod@blastradius.com>]

> Within an enterprise you may want to disallow further 
> specialization of a set of types. That is, one key benefit of 
> specialization is that it can be *controlled*. Thus it 
> follows that you should have the option of disallowing it, 
> as well as allowing it.

Would you buy the argument that a type should be able to control the
addition of other attributes if and only if the type can control other
aspects of specialization? E.g. if a type can declare that it is or is
not specializable or allows for content model tightening? Do you agree
that controlling attribute specialization is just a small part of it? If
so, then I would defer it until the whole issue of specialization
control is covered holistically.

> This is recognition of the fundamental problem of recognition posed by

> the use of the no-namespace namespace--namely, you have no reliable, 
> general way to tell what application the attributes correspond to.
Thus 
> it seems like it might be useful to disallow the confusion by
requiring 
> non-DITA-defined attributes to be qualified.

We could make that rule globally. Or we could give specializers control
over it as part of the specialization control issue. I would say that
attributes belong to the same namespace/domain as the elements they are
on unless they use a namespace prefix explicitly.

 Paul Prescod