Discussion of ‘the Semantic Web’ and ‘Web Services’ rarely dies down, but it seems like there’s been more than usual of late. As analysts yet again predict that this year will the “The Year of Web Services” more people are beginning to agree that such hype isn’t making any real impact on the use of service-based architectures.

One of the better comments of late is Danny Ayers’ call to start seeing web applications as “features of the web” rather than standalone entities. The Semantic Web will never evolve if we merely see web services as ways to escape writing certain pieces of code ourselves; we need instead to be trying to grasp at how the whole might work.

As my previous post indicated, I’m working on some ideas for some local websites designed to develop community and support local businesses. While there’s certainly a need for distinct interfaces to the data–to enable people to focus on their particular interests, or so that multiple organizations can have a sense of ownership–each interface will be considerably enhanced if the sites are geographically aware and interoperable.

If, for instance, I am going to a concert downtown I’d like to know what other events are happening nearby or where to eat. Rather than having to go and perform extra searches elsewhere, it would be useful if the site presented me with some of that information. And rather than having to have one organization keep track of all the information, it would be far better to have everything provided in a machine digestible format, with the different sites querying one another (either on-demand or regularly) and a simple way of plugging in other information sources.

That’s why Grand Rapids WiFi is now publishing as much data as possible in RDF/XML, and other sites will do likewise. I suspect that the development of the web as a platform will follow a similar route to the adoption of standards-based design: more and more of us started migrating/building our small sites that way, pressure built and suddenly the big guys were getting in on the act. To be honest, I can’t remember whether the analysts were in on that act, but I also can’t remember that mattering most of the time.

Recommend this post:

Looking to the future of the Grand Rapids WiFi site, I hope to see it become part of an integrated set of local websites supporting and promoting community development and local business. Geolocation seems to be the topic du jour, and while the site has for several months featured geodata about all of its locations the time seemed right to develop it further.

Today I’ve been adding RDF representations of almost all the data on the site. I’ve extended the RDF descriptions of each location to list all the comments on that location. The vocabulary for that is one I found over at FilmTrust and it means that almost all the useful content of the site can now be represented using RDF.

To tie it all together I’ve added an RDF/XML index of the site so that agents can easily get a complete list of WiFi locations listed on the site. Obviously a spider could have worked its way through the site and found all the data, but this removes a hurdle.

It’s remarkably easy to add RDF representations of data on a database-driven site such as this one. There was some tweaking to do, but mostly the work was constructing and validating a new set of templates. Now I need to start harnessing the new power.

Recommend this post:

For the past couple of weeks I’ve been trying to get the Redland RDF toolkit installed on my powerbook. I initially tried installing it using the vanilla tools, but kept running into the error:

ld: common symbols not allowed with MH_DYLIB output format with the
-multi_module option
/usr/local/mysql/lib/libmysqlclient.a(my_error.o) definition of common
_errbuff (size 512)
make[2]: *** [librdf.la] Error 1

I upgraded a few of the standard tools (such as libtool) using the latest sources from the GNU website, and Raptor and Rasqal compiled without a hitch. But Redland still threw that same error and google turned up little in the way of help.

Having googled over and over again, it seemed the only way anyone had managed to get Redland installed on OS X was using either fink or darwinports, so I installed fink, but once again ran into the same error.

In the end, a solution came in removing the MySQL instance that I had installed using the binaries on the MySQL website and switching to the fink version of MySQL.

Removing MySQL allowed me to install Redland, and build the perl language bindings, but the python bindings kept throwing the error:

gcc -g -O2  -I/usr/local/include -I/usr/local/include -I/usr/local/include \
-I/sw/include/db4  -g -O2   -Wl,-F. -Wl,-F. -bundle Redland_wrap.so \
`redland-config --libs` -o Redland.so
ld: Undefined symbols:
_PyArg_ParseTuple

followed by a long list of other python symbols. My system seemed to be using the fink-installed python rather than the system default, so I tried configuring with:

./configure --with-python=/usr/bin/python

Which seems to have done the trick. At least, ‘import RDF’ works from the system python.

Recommend this post: