Page:Aaron Swartz s A Programmable Web An Unfinished Work.pdf/48

From Wikisource
Jump to navigation Jump to search
This page has been proofread, but needs to be validated.

36 5. BUILDING A PLATFORM: PROVIDING APIS

    >      ],
    >    }

And if your site let people update book pages, you could imagine supporting PUT requests on this URI that allowed people to submit an updated version of the JSON object. You’d parse it and then execute the update.

Or, if you just let people comments on books, you could let them POST simple JSON data to the same URI that comments are normally posted to.

In fact, if your really wanted, you could just let them POST form data and parse it the same way as you would input from web browsers. Then you could let them know success or failure via HTTP error codes—a 500 error would let them know it failed, while a 303 See Other redirect to the page itself would let them know they succeeded. When they followed the redirect and grabbed the page, it too could content negotiate to JSON.

    ***

Alright, now it’s time to talk about a touchy subject. I’ve been holding off on this, but at some point it becomes unavoidable. Yes, I’m afraid it’s time to talk about RDF.

You see, all this JSON stuff is great for writing little scripts on clients that talk to other scripts on servers, but it leaves something to be desired when working at Web scale. It’s hard to imagine, for example, building particularly useful tools that work across different JSON APIs, the way web browsers work across all different kinds of HTML pages. Each JSON API has its own internal representations and conventions and protocols, which means you need to write special code to deal with each different one.

That’s where RDF comes in. The idea behind it is simple: what if we had a format that did to data what HTML did to documents—provide a single, consistent representation for them that supports the hypertextual nature of the Web. That probably makes no sense, so let’s look at some examples.

RDF documents are quite simple—they’re made up of “triples,” simple sentences with three parts: a subject, a verb (called a predicate), and an object. Let’s take a bit of our example from before, namely that the book with ID 3j7is has the title “The ABC book”—in RDF, the subject would be “3j7is”, the verb “title” and the object the string “The ABC book”.

Only RDF is meant to work at webscale, so instead of fuzzy-wuzzy terms like “title”, everything’s a URI. As in: