Hi Jacco and Stefano and all,
----- Original Message -----
From: "Jacco van Ossenbruggen" <Jacco.van.Ossenbruggen_at_cwi.nl>
To: "Stefano Mazzocchi" <stefanom_at_mit.edu>
Cc: <general_at_simile.mit.edu>; <lloyd.rutledge_at_cwi.nl>
Sent: Friday, April 22, 2005 4:24 PM
Subject: Re: AW: Comments on Fresnel and your ISWC draft
> Stefano Mazzocchi wrote:
>
> >
> > this goes back to the styling issue: I think that fresnel should be a
> > bridge between a graph and a tree... and once you are in 'tree world'
> > you can use whatever you want to present that, as there are many
> > technologies and best practices already.
> >
> Good point. Maybe you'd like to have a look at earlier work we did.
> This is currently going through a long review/editing/publishing cycle
> of the IJWET journal (see below for bibtex ref). AFAIK it is still not
> published :-(
>
> An old pre-print, however, is available online:
>
> http://ftp.cwi.nl/CWIreports/INS//INS-E0303.pdf
>
> The approach is what I call a minimalist approach to the RDF graph to
> XML tree problem. The only thing I did is write a simple java class
> (about 1 page of code) that implements an XSLT extension using the Xalan
> XSLT API. The java class sends an RDF query to a query engine (in our
> case, Sesame) and returns the answer as an (XML-encoded) table of
> variable bindings. This is returned to the XSLT style sheet, and you
> are back in "tree world" and can do further processing in XSLT & CSS.
>
> Advantages:
> - you can mix and match the languages. Things we do all the time is
> using XSLT parameters and variables in RDF queries, select different
> XSLT template rules based on results of RDF queries, style RDF query
> output roughly by XSLT rules and fine tune by using external CSS style
> files, etc.
> - Easy for people that already use XSLT. For example, Noadster was
> build by inserting a single SeRQL-extended XSLT style sheet in an
> already existing XSLT-based Cocoon pipeline.
> - minimal amount of code required, just the simple java class that glues
> the XSLT and Sesame engines together.
> - once SPARQL is a W3C REC you can replace SeRQL by SPARQL and the whole
> solution only requires a combination of two W3C RECs.
>
> Drawbacks:
> - the query syntax becomes an ackward mix of XML, XSLT and the RDF query
> language. As a result you have to escape the angle brackets thay many
> RDF query languages use, for example in the <http://example.com>
> notation. Serql's use of {} conflicts with that of XSLT, etc.
> - People who hate XSLT will hate this even more: not declarative, hard
> to read syntax, etc.
>
> The drawbacks also explain our interest in Fresnel: I hate the idea of
> further extending Noadster using this XSLT/SeRQL combination. It is
> great for Q&D fast prototyping of small demos, but I'm affraid it
> doesn't scale to a real, customizable RDF display framework. But I
> think it does show that Stefano is right in saying that the graph to
> tree problem is the main issue. Once your in tree world, I would stick
> to simple CSS wherever possible and use XSLT for the more complex stuff
> if needed. If Fresnel would only give me a declarative solution for the
> graph-to-tree problem (a transformation which I could, for example, plug
> into Cocoon), I would be very happy!
>
Fresnel gives you this declarative solution by using Fresnel lenses together
with the RDF-specific part of the style vocabulary (fresnel:label and
fresnel:value). Using this parts you can generate an XML tree which you can
feed into Cocoon for further processing. Valid solution of the problem, but
having the drawbacks you descibe.
These drawbacks and the idea of expressing general display rules which are
valid across different visualization paradigms (for example, display all
persons that are older than 30 years using a special color) lead to the idea
of the abstract box model (supported by Emmanuel and me) which is still
under discussion as you can see from Stefano's mail. We think attaching
styling instructions to an abstract model which is interpreted according to
a browser's visualization paradigm and output format is a cleaner solution
than attaching the styling instructions in the XML transformation process.
But it is also true that you said that people might hate either one or the
other solution depending on their background. XML people like to write down
display knowledge in a procedural manner as XSLT transformation steps. RDF
people like to write it in a more general, declarative way.
We tried not to exclude either group when designing Fresnel. A set of
Fresnel styles and lenses can be used by an RDF visualization tool for
generating input for a XSLT pipeline by interpreting only the lenses and
fresnel:label/fresnel:value properties but ignoring the box model
properties. Another tool might want to take advantage of the box model by
using the display knowledge encoded using the box model properties and the
additonal content properties.
Chris
> Jacco
>
> Bibtex for the IJWET article:
> _at_article{ins:ijwet04,
> volume = {1},
> note = {to be published, based on
> http://ftp.cwi.nl/CWIreports/INS//INS-E0303.pdf},
> author = {Jacco van Ossenbruggen and Lynda Hardman and Lloyd
> Rutledge},
> journal = {International Journal of Web Engineering and
Technology},
> title = {{C}ombining {RDF} {S}emantics with {XML} {D}ocument
> {T}ransformations},
> number = {4},
> year = {2005}}
>
Received on Fri Apr 22 2005 - 17:03:59 EDT