- From: John Cowan <cowan@mercury.ccil.org>
- Date: Thu, 6 Sep 2012 03:22:26 -0400
- To: James Clark <jjc@jclark.com>
- Cc: public-microxml <public-microxml@w3.org>
James Clark scripsit:
> > I think (a) here can be divided into (a1), the content of the document,
> > and (a2), the embedded metadata. PIs (and the DOCTYPE tag constitute
> > the latter. They are not in the data model but are a bag on the side.
>
> I can see this more for PIs in the prolog (where the position of the PI is
> not really important). It feels much less natural for PIs in content, where
> the position is crucial.
Agreed.
> And even if you separate out (a2) from (a1), you still need a data model
> for it. If you put a gun to my head, I could live with PIs with start-tag
> syntax in the prolog, and then provide a data model for the content (what
> we have now) and a data model for the metadata (ie a list of PIs).
I agree with that too. There is no law that says the data model most be
in one piece, so to speak: the LMNL data model has a sequence of atoms
(content characters are atoms, but there are other atoms too) and then
a list of ranges over those atoms.
> It doesn't make sense to me to have PIs in the prolog like xml-stylesheet
> and say that these are only for markup sensitive applications.
No, it doesn't. But I wasn't saying that, I was saying they were
post-data-model decorations that can be added for XML compatibility
and are thrown away again (or reported as second-class citizens) when
the MicroXML is parsed. But I like the idea above better.
> Having PIs in content being for markup sensitive applications (as in
> oXygen) makes sense, but I don't think a special syntax is necessary,
> because I think such applications could easily adapt to using comments with
> some sort of prefix (as in Javadoc and similar things).
I reluctantly agree.
--
Overhead, without any fuss, the stars were going out.
--Arthur C. Clarke, "The Nine Billion Names of God"
John Cowan <cowan@ccil.org>
Received on Thursday, 6 September 2012 07:22:52 UTC