So tomorrow morning it's off to Alexandra Palace for the start of Hack Day London. My wife is raising an eyebrow as I ready my sleeping-bag, and she may have a point...but I'm secretly quite looking forward to hanging out with a lot of smart people, hearing about their good ideas, with no-one around to tell me what time to go to bed.
For myself, I'm hoping to get involved in projects relating to metadata and rich-clients. I'm sure there will be people working in this area, but if not, I've got plenty of new things we've been experimenting with that I might be able to get others to take a look at.
RDFa Parsing
Perhaps the most interesting is the use of our RDFa parser within a blog or other document. By adding metadata to elements in a document you are providing a hook which gives you more information about some item. Onto this hook you can 'hang' further functionality or more information.
To illustrate, let's say I have a cookery blog on which I mentioned Canteen Cuisine by Marco Pierre White. Given that I probably want to make some money from the site with my Google Ads and reseller links, I will most likely place a link to a site like Amazon around the words "Canteen Cuisine":
I found a good recipe in <a href="http://www.amazon.com/gp/product/0091808189?ie=UTF8&tag=escuelerie-20&linkCode=as2&camp=1789&creative=9325&creativeASIN=0091808189">
Canteen Cuisine
</a> which you really should buy so that I get a kick-back.
Obviously this satisfies my commercial yearnings, but it doesn't really contribute to the great 'mass' of semantic information that others might find useful. It doesn't actually tell you which book is being referred to, only that there is a link to a page on Amazon. It would be far better--and easier, as it happens--to simply tag the book with an ISBN number, and leave the creation of links to Amazon, or Barnes and Noble, to some other layer of behaviour.
The key is to use RDFa to indicate a unique identifier for the book, and also to indicate an object type, i.e., to indicate that the item we're dealing with is really a book:
I found a good recipe in <span xmlns:bib="http://somebig.org/" about="urn:ISBN:0091808189" class="bib:book">
Canteen Cuisine
</span> which you really should buy so that I get a kick-back.
If you know RDF then you'll understand that the generated triples are:
<urn:ISBN:0091808189> rdf:type <http://somebig.org/book> .
If you don't know RDF don't worry; the main point is that we've added mark-up that says nothing more than:
"urn:ISBN:0091808189" represents a book.
But that simple fact is pretty useful, since now we could use the ISBN number to go and get more information from Amazon, and then display the data in a tooltip, or add links so that a reader of the cookery blog could buy the book or add it to their wish-list.
Before we look at how we get that data, it's worth pointing out that Amazon does offer the Quick Linker widget for TypePad users, but you'll see that the mark-up looks very strange:
<a type=amzn asin=B000012345>I love this item</a>
In some ways it's similar to what I've shown with RDFa, where we use a small amount of mark-up as a hook to generate richer functionality, but in their case it's an Amazon-specific solution. A key design goal of RDFa is to provide a generic solution that would work with all types of data.
RDFa action handlers
Once the RDFa parser has 'found' items in the page it then does something with them. Up until now all of our experiments have been to simply display data in a different way, such as showing FOAF information as a card with the person's picture, or event information as a pin on a map. The Operator Firefox extension (which now supports RDFa as well as microformats) takes a slightly different approach and allows the user to do things with the data, such as add an hCard to their contacts database.
The Amazon book experiment is a first step towards combining these approaches. The action handler that gets run when a book is found actually goes to get further data about the book from Amazon, before it shows a panel that contains an image of the book and its title. This means that we don't need to know any URLs for the book on Amazon or where its images are located, since they are derived in the action handler; all we need is the ISBN number.
Yahoo! Pipes
But an interesting twist here is that we don't go directly to Amazon to get the data, but instead go via Yahoo! Pipes. There are many reasons for doing this--not least that it fits well with my ideas about skimming--but the main one is simply because Pipes gives us JSON data for any feed we can access, which works nicely with the model we're building here. By running all of our data requests through Pipes we can provide a single format to our RDFa processor, even if we decide to change the source of our data at some later point, or add further services into the pipe to add more metadata.
The pipe that we've defined for Amazon actually takes the URI of the ISBN number (such as urn:ISBN:0091808189) rather than just the ISBN number, because at some point I'd like this to be a proper RDF query. So I'm actually saying, "give me everything you know about this URI", and whilst that URI currently represents a book, in the future the same query might be able to return information about URIs that represent people, ships, planets, and so on.
What's next?
Everything we've been working on so far is about providing a solid widget framework that can be used to enhance any page on any site, in a simple and consistent way. I'm sure that at Hack Day there will be people working on ideas to do with enhancing blogs, ideas related to using the semantic web, microformats (whether in the traditional or RDFa style), and many other things besides, and so I'm really hoping that this will be an opportunity to move some of this work on, in the context of real applications that people want to build.
