Hi Erik--
Were constraints an off-the-list 1.1 item? Too bad we couldn't get to
them last time.
In any event, having just implemented a customization myself to
eliminate many unneeded elements and attributes from the content model
- but in the fashion of a strict subset, so that full compatibility is
assured (well, except for a few renamings) - I can attest to how
valuable this systematic solution might be - at least in the document
type shell file
On the other hand, since my document instances retrieve their schemas
from a webserver, I'd probably resort to customization anyway on the
Mod and Grp files, to eliminate the performance hit of retrieving and
parsing so many unused elements and attributes when opening a document.
--Dana
Erik Hennum wrote:
Hi, Eliot:
Good design pattern -- probably even be a best practice for DITA 1.1
specializations.
I do wonder whether the applicability of this pattern to many cases
means, paradoxically, that we shouldn't introduce a specific element
for the individual case. For instance, the pattern also applies to a
context of mixed text with blocks (as in <section>, <p>,
<entry>, or <itemgroup>).
The other consideration might be that, if we can come up with an
acceptable implementation to the constraints proposal, we could apply
the constraint to the content model and remove the need to specialize:
(Assuming we can come up with a constraints implementation that can be
detected at runtime to determine interoperability, that packages
constraints in reusable modules, that applies constraints to
specializations, and that doesn't require an inhumanly complex design
pattern for the DTD and XSD implementations.)
Hoping that's interesting,
Erik Hennum
ehennum@us.ibm.com
"W.
Eliot Kimber" <ekimber@innodata-isogen.com>
At least with DITA 1.1, where we can't add attributes, a
reasonably
common pattern is this (using the base types, which would normally be
specialized for a specific application):
<ph>
<data/>
<ph>Text content of the phrase</ph>
</ph>
...
Given the general utility of this pattern, it seems reasonable to
provide the "text" specialization of "ph" as a base type so
specializers
don't have to repeatedly re-invent it.
|