Re: lenses

From: Eric Hanson <>
Date: Tue, 26 Apr 2005 20:09:02 +0000

Chris Bizer ( wrote:
> >> So I guess I'm arguing that lenses are valuable and you're using them
> >> totally right, but you might save some work and make them more
> >> generally useful by just defining them as RDF serializations of
> >> queries, be it SPARQL or other languages.
> Yes, your are generally right. Lenses are a domain-specific RDF query
> language.
> They are less expressive than SPARQL (which is good for our domain), but
> provide ordered results, what SPARQL doesn't do.


Ok, so RDF is an unordered infoset. It seems like putting the
horse before the cart to rely on the query mechanism to do the
ordering. In fact, why is ordering important at all, except for
at that very last step, the display? One motivation, it allows
transformation with languages such as XSLT that rely on
ordering. But it seems that a conversion from RDF to an ordered
form or visualization format should be a mapping, not a
sequential transformation.

I guess I'm thinking in terms of the pipeline approach [he says,
shamefully soliciting a nod from stefano :-]. The options I see

1. Retrieve data, order it, branch on missing info etc w/ lens
2. Transform it (by hook or crook) to XHTML (etc)

1. Retrieve data with a SPARQL pipe
2. Order it with a ordering pipe
3. Transform it with a transformation pipe

1. Retrieve data with a SPARQL pipe
2. Use a mapping pipe that takes an unordered subgraph and
maps it to a template.

The third option set seems compelling to me. It relies on a
(admittedly somewhat half-baked) mapping to a template, where
the mapper pipe plugs values from the unordered result set into
the order perscribed by the template.

> Ordered results are also important in the security context for being able to
> apply signatures. Jose was talking about this idea in his W3C tech plenary
> talk. See:
> There is a EC-project about SemWeb security in the funding-pipe and if we
> get money there, we will definitively look closer at these issues.
> But for now I think lenses are a domain-specific query language designed for
> being used in the context of visualizing RDF data and we should just leave
> it this way and don't try to fulfil requirements from other domains.


So, sorry for kind of jumping in here, but I'm keenly interested
in this issue. For a little background, I'm a graduate student
at Portland State University, working on SW stuff mostly as a

I worked on a now defunct project called Typekit[1] that was
aimed at being a general vocabulary for describing supporting
resources and associating them with XML Qnames. Supporting
resources are things like (but not limited to) UI components for
display, editing and creation. Adapting the underlying ideas to
the RDF model is what I'm working on now and it seems like there
could be some overlap here.


Received on Tue Apr 26 2005 - 20:08:14 EDT

This archive was generated by hypermail 2.3.0 : Thu Aug 09 2012 - 16:39:18 EDT