In message <199704101041.GAA00273@localhost> David Megginson writes:
> Peter Murray-Rust writes:
>
[...]
> > - are AFs part of HyTime or more generally part of SGML?
>
> Architectural forms were originally specific to HyTime, but in the
> forthcoming Annex 1 of the standard, they have been generalised --
^^^^^^^^^^^^
Which? 8879?
> HyTime is one base architecture, but DSSSL is another, as is the
> Canadian GILS project.
>
> > - does XML support AFs without further revisison?
> > - if not, what has to be added? or is it possible to build this
> > into tools without breaking the spec?
> > - if tools have to be built, is the processing required at
> > parser level?
>
> There are different ways of dealing with AFs. James Clark's SP
> library can do special, smart AF processing, where it actually parses
> the document as if it were an instance of the base architecture rather
> than of the derived architecture, but that is not always necessary (or
> desirable). XML-based tools can simply look for the <?ArcBase ...?>
> processing instruction and the associated notation declaration, and
> then use the attribute values for processing.
I'm getting the impression that this is at *processor* level, not parser.
I.e. we might need an AFprocessor tool in our Java toolkit? Comes between
the parser and the application?
>
> NXP-based apps, for example, should be able to handle this, but
> DTD-less parsing will not be possible unless the attribute values are
> included in the document instance itself (I consider this a feature
> rather than a bug).
This means they are added to the declaration subset or to the pointers
to the entities?
>
> The one short-coming is that XML does not support data-attributes, so
> it will not be possible to customise the AF support -- you'll have to
> stick with the defaults.
This sounds reasonable for a first step :-)
>
> > - do the components belonging to different DTDs have to live in
> > separate files or can they be separated within a single
> > file (e.g. by NOTATIONs)?
>
> They should be stored in separate files (etc.), just like public
> entity sets.
>
> > - if someone decides they need AFs are they easy to implement?
> > (for my requirements it looks like an aliasing mechanism
> > would suffice.
>
> This _is_ an aliasing mechanism, of sorts.
Excellent. I suspect this is likely to be critical when people start reusing
tagsets. Other opinions?
P.
-- Peter Murray-Rust, domestic net connection Virtual School of Molecular Sciences http://www.vsms.nottingham.ac.uk/