W3C home > Mailing lists > Public > public-lod@w3.org > April 2009

Re: Making RDF / LinkedData trivially browseable - thoughts?

From: Keith Alexander <k.j.w.alexander@gmail.com>
Date: Sat, 04 Apr 2009 13:10:37 +0100
To: "Daniel O'Connor" <daniel.oconnor@gmail.com>, "Hugh Glaser" <hg@ecs.soton.ac.uk>
Cc: "Kjetil Kjernsmo" <kjetil@kjernsmo.net>, "public-lod@w3.org" <public-lod@w3.org>
Message-ID: <op.uru2rze063ayaz@polar-bear>

I strongly agree with "forget about the XML toolchain for RDF". There are  
so many different ways an RDF graph can be serialised as RDF/XML that I  
think it would be hard/dull to write an XSLT that you could be really sure  
would work on all RDF/XML. Your web developer might produce valid RDF/XML  
and think there was something wrong with it if your stylesheet didn't work  
on it.

On Sat, 04 Apr 2009 10:55:24 +0100, Daniel O'Connor  
<daniel.oconnor@gmail.com> wrote:

> I'd have the same arguments for enabling some low level kind of browsing  
> of
> data as I would for publishing RDFa - its easy, interactive, and
> understandable; while still doing all of the neat stuff with linked data.

Why publish RDF/XML with an XSL stylesheet to make it into generic-looking  
HTML, when you could publish RDFa which you can style and script however  
you like?

> The learning curve I sort of see is for an average developer who stumbles
> across a dataset:
>    1. Its really easy to see how this data is connected. I might just  
> right
>    click, and view source, and look what's under the hood.
>    2. Hey, this is all XML right? If I were going to build an application
>    (ie, a mashup), I could do a whole bunch of xpath and http get  
> requests!

Can you think of a way to encourage them to skip step 2 ? It will only  
lead to pain and confusion ...

>    3. I've just built three or four mashups; maybe there's a better way  
> that
>    xpath and http get? What's the underlying RDF stuff all about? What  
> are all
>    these identifiers about?
>    4. Oh: wow; RDF is neat; the next time I have a data set or plan to  
> make
>    my own API, I might as well just publish it as RDF/XML / linked data.

Maybe it would be better if they didn't publish it as RDF/XML (or at least  
not only as RDF/XML), but as RDFa, or turtle, or RDF/JSON .
I know RDF/XML is probably still the best supported by RDF tools and  
applications (are there any surveys of the formats supported by RDF  
software?), but if the aim is give mainstream web developers a smooth path  
into linked data, don't these other formats do it better, and avoid the  
"marketing" problem of making RDF look like XML?

Turtle is most readable and compact (and tool support is quite good).
RDFa is most human-friendly, and you can follow links, see images, run  
media, etc, in browsers.
RDF/JSON is most (I think) transparent for scripting with. Simple uses  
don't require special RDF libraries - just a json parser.

Not that it's very hard to provide different serialisations and do  
content-negotiation, so you don't need to pick only one.
Another option could be a few lines developers could add to their  
/.htaccess file to redirect to a converter service like http://triplr.org  
or http://morph.talis.com if the user is requesting a serialisation you  
aren't supporting.

>> Tell your "regular web developer guy" that all he needs are the URIs for
>> the thing he is concerned with, so he can resolve them, and then he can  
>> do a
>> "Mash-up/Mesh-up" with the data he gets back to offer lots more data to  
>> his
>> customers, from a very wide range of sites.
> No one is doing that though!

http://revyu.com (for one) does this - retrieving your foaf file if you  
provide one, and mashing it up with the application's info about you, on  
your profile page.

http://tinyurl.com/d4l8w3 is an easy demonstration of mashing up linked  
data. I have 2 different documents with data about me, both using the same  
URI to identify me - they can be easily merged into a new document  
describing me with all the properties from both documents.

http://simile.mit.edu/babel will also do this kind of thing (merging data  
 from multiple documents). It also does it with bibTex, but one of the  
values this demonstrates in using RDF is, by using URIs, the properties of  
a resource can be spread across multiple documents, and the data is not  
constrained to any particular domain.

> You aren't seeing a lot of the simple, easy tricks that come with
> data published as linked data being talked about.
It would be cool to see more of this kind of thing. I bet people on this  
list know lots of cool tricks and tips - can you suggest how they could  
get more exposure? What kind of thing do you think mainstream developers  
would be most interested in seeing?

> Its not easy to get a triplestore up and running in 5 easy steps, its not
> easy to see the benefit of modelling your data, its not easy to see why
> linked data and cool uris are better than printing out a whole bunch of  
> from your web application.

When I started getting interested in RDF, I found it pretty easy to set up  
a triple store in a LAMP environment with ARC (http://arc.semsol.org/) in  
less than 5 steps. From there, the benefits of RDF made themselves  
apparent quite quickly :)

And JSON is still good if it is RDF/JSON :) (especially with a JSONP  


Keith Alexander

Received on Saturday, 4 April 2009 12:11:30 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 16:20:46 UTC