forrest-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jeff Turner <je...@apache.org>
Subject Re: [VOTE] Adopt site: linking
Date Wed, 08 Jan 2003 16:25:21 GMT
On Wed, Jan 08, 2003 at 04:28:16PM +0100, J.Pietschmann wrote:
> Jeff Turner wrote:
> >It seems RDF has the advantage of being representable as almost-plain-XML,
> >whereas XTM wants a custom format.
> What custom format? Last time I looked it was XML too.

I phrased it badly.  XTM *is* the custom format.

>From my limited knowledge of the formats, I'd say the XTM structure gives the
datamodel (topics, associations) precedence over the data (files, dirs),
whereas with RDFs abbreviation rules allow one to express data (see XML below)
that 'incidentally' forms a data model (subject, predicate, object).

Though, I lack the imagination to guess what Forrest could actually DO with RDF
triples or TM topics/associations.  I have the feeling there's a "killer app"
waiting to be discovered here.  Until this becomes clearer, I'd like to keep
the idiot-proof site.xml format.

> >  <folder rdf:ID="/" about=".">
> >    <file rdf:ID="index"
> >      dc:Source="index.xml"
> >      dc:Description="Forrest home page"
> >      dc:Creator="Various"
> >      dc:Format="text/xml"
> >      href="index.html">
> Cool!

You snipped the best bit! :)  How with RDF:
1) it doesn't matter if the information is an attribute or element (much like RNG)
2) if an element, the data could be a Bag, Sequence, or in this example, an
either/or 'Alternative' object[1]:

      <dc:Title>
        <rdf:Alt>
          <rdf:li xml:lang="en">Index</rdf:li>
          <rdf:li xml:lang="de">Das Index :)</rdf:li>
        </rdf:Alt>
      </dc:Title>

> Now for a more complex problem:
> There is a directory "design" containing a bunch of
> xdocs which may have links to each other, let's say
> there is a <link href="site:svg-embedding"> somewhere.
> No problems for the website, but what if, as Keiron
> described, the idea is to generate a PDF booklet
> containing all the design stuff? This means you may need
> different mappings for the same link destination,
> dependent on some context.

*fzzzt* brain overload..

I'm sure there is a Grand Unified Theory of how Forrest linking could
accommodate this and similar scenarios.  I'm also pretty sure that
site.xml is not indicative of the right long-term approach, because it
which mixes the tasks of

1) naming (addressing) content (the XPath address)
2) associating metadata with content (eg @title)
3) ordering and categorising content ('index' before 'download', and both
   in 'About' category)
4) Defining 'views' of the above. 'index.html' could be considered a
"view" of 'site:index'.  What Keiron is asking for is a View composed of
multiple 'site:' nodes.  How do we circle a bunch of nodes and say "these
belong in view X"?  Links inside X must be treated differently to those
outside X.

Answers on a postcard, please.

--Jeff


[1] http://www.w3.org/TR/1999/REC-rdf-syntax-19990222/#containers

> J.Pietschmann
> 

Mime
View raw message