RE: [dita] Proposal: Update foreign topic to include processing info forgeneralization/specialization use case

[Robert D Anderson <robander@us.ibm.com>]

Thanks, Paul.

I do have one other thought on this -- rather than trying to hide foreign
content during generalization, what if we export it -- essentially, turn it
into an imbedded object? Others have pointed out that foreign content is
really an inline object, so this seems to make sense architecturally. We
could pretty easily turn this:
<math class="+ topic/foreign mathml/math ">
  mathml content here
</math>

into
<foreign class="+ topic/foreign mathml/math ">
  <object data="imbed1.xml" type="DITA-foreign"/>
</foreign>

Some additional details on this:
*  The object element may already be used for alternate content. That would
not be exported - so you end up with one object pointing to the foreign
content (type="DITA-foreign" or some equivalent), and one object with the
alternate content.
*  Foreign content may use several main elements -- such as
<foreign><this/><that/></foreign> -- so the exported file should have a
root <foreign> element to remain valid.

I think the basic approach of exporting the info and using <object> is a
lot more elegant than processing instructions, both from a DITA perspective
and from a general XML perspective. It allows me to edit my topics with
mathml, and then generalize and send them to somebody without mathml for
further editing/processing. It also doesn't use any kludges to hide XML as
if it was not XML.

Does this seem like an approach worth pursuing?

Thanks-

Robert D Anderson
IBM Authoring Tools Development
Chief Architect, DITA Open Toolkit
(507) 253-8787, T/L 553-8787

"Grosso, Paul" <pgrosso@ptc.com> wrote on 10/18/2006 08:45:31 AM:

> Actually, I'm okay with 1, 3, or 4, but 4 is what I was
> suggesting on the telcon and would be my preferred path.
>
> ---
>
> What are the options with XML Schemas?
>
> Am I right to think that--as long as the foreign content
> is namespaced--we can get things to work by giving the
> foreign element the appropriate content model (i.e., one
> that allows any namespaced element)?
>
> ---
>
> As far as DTDs, I think we could give the foreign element
> an ANY content model and then all the user would have to
> do is insert declarations such as
>
> <!ELEMENT my:foreign-element ANY>
>
> for each foreign element (without explicitly adding references
> to these elements in any content model), and things would work.
>
> That doesn't change your point 1 substantively, but it does
> reduce "using you foreign DTD module" to merely adding a
> simple declaration for each of your foreign elements making
> it much simpler to "share" with others.
>
> paul
>
> > -----Original Message-----
> > From: Robert D Anderson [mailto:robander@us.ibm.com]
> > Sent: Tuesday, 2006 October 17 13:16
> > To: dita@lists.oasis-open.org
> > Subject: RE: [dita] Proposal: Update foreign topic to include
> > processing info for generalization/specialization use case
> >
> > It seems like - if we support the foreign and unknown element
> > - we have
> > these options when using DTDs:
> >
> > 1. We state that, if you use the foreign or unknown elements,
> > you have no
> > way to share topics with others that do not use your foreign
> > DTD module.
> > This would be the only way to make a DITA document that
> > cannot be taken
> > back to the topic type that represents a
> > least-common-denominator, because
> > you can never share with somebody that does not use this module.
> >
> > 2. We define some way to hide the information when
> > generalizing. Groups can
> > go back to the most common form; if the common form does not use the
> > foreign specialization, then they use the alternate content
> > they (may have)
> > defined with <desc>. So far the only real proposal for hiding the
> > information is with a PI, which has already been declared a kludge.
> >
> > 3. We state that foreign content is simply discarded during
> > generalization
> > - you can't bring it back if you're going to a format that
> > doesn't support
> > it.
> >
> > 4. . We leave it up to each implementation. This means that if you
> > generalize with one program, other programs may not be able
> > to bring it
> > back. Currently the spec defines how to generalize, which
> > means that all
> > implementations should be able to work with
> > specialized/generalized content
> > in the same way.
> >
> > Does anybody see any other options? Paul, it sounds like
> > you're suggesting
> > option one -- does that sound right?
> >
> > Thanks-
> >
> > Robert D Anderson
> > IBM Authoring Tools Development
> > Chief Architect, DITA Open Toolkit
> > (507) 253-8787, T/L 553-8787
> >
> > "Grosso, Paul" <pgrosso@ptc.com> wrote on 10/17/2006 11:48:36 AM:
> >
> > >
> > > > -----Original Message-----
> > > > From: Eric Sirois [mailto:esirois@ca.ibm.com]
> > > > Sent: Tuesday, 2006 October 17 09:32
> > > > To: dita@lists.oasis-open.org
> > > > Subject: [dita] Proposal: Update foreign topic to include
> > > > processing info for generalization/specialization use case
> > > >
> > > > The foreign content proposal (#35) for incorporating foreign
> > > > content in
> > > > DITA did not have a use case that covered what would
> > > > happen when an DITA document contain foreign content are
> > > > generalized the
> > > > content is not longer valid, when using DTDs for
> > > > validation, because the base DTD does not include the foreign
> > > > vocabularies.
> > >
> > > When using DTDs, if it is desired for the generalized
> > > content to be DTD-valid, then the foreign content elements
> > > should be added to the DTD.
> > >
> > > > I propose that we hide the foreign content via a processing
> > > > instruction in
> > > > the generalized document. It gets us past the validator
> > > > issue. The content could then be re-instated during the
> > specialization
> > > > process.  Hiding in a comment was discussed, but there
> > > > is no guarantee that tools keep comments in the documents.
> > > > Also, it makes
> > > > reinstating the foreign content during specialization
> > > > much more difficult.
> > >
> > > Putting what should be markup as characters in either a
> > > comment or processing instruction (PI) is quite distasteful
> > > to me.
> > >
> > > Also, many tools toss PIs too--not ours unless such is
> > > explicitly requested, but others do.
> > >
> > > Note, your generalization process will need to get tricky
> > > if the content you are trying to hide in a PI contains
> > > a "?>" string.
> > >
> > > Finally, this all sounds too kludgy and implementation based
> > > to me for a specification.
> > >
> > > >
> > > > In the case where <foreign> contains an <image>, <object>, or
> > > > <desc>, they
> > > > should not be hidden within the processing instruction.
> > > > All content within <unknown> will be hidden in a processing
> > > > instruction.
> > >
> > > paul
> >