Similarity Web 2.0: Remaking My First Mashup, 5 Years Later

One fateful summer five years ago, I accidentally stumbled upon the world of Web APIs - starting with the Amazon E-Commerce API. I was supposed to be working on a research project, but I couldn't resist the potential of the API. Amazon's catalog & functionality at my fingertips? And I could do whatever I wanted with it? I just had to play with it.

So I made my very first mashup: the Similarity Web. It's an app that uses the Amazon ItemLookup & SimilarityLookup API calls to let you find a book & visualize the web of books similar to it. It was a fun little app, and it even made me some money in referrals - especially after marketing guru Seth Godin linked to the "Purple Cow" web in his blog. Unfortunately, as it happens, Amazon deprecated the version of the ECS API that I was using, and the app stopped working. It's been on my to-do list for the last few years to revive it.

Well, thanks to my former employer, I found an excuse to do just that. Google is holding a contest for free I/O tickets, and a few Fridays ago, they held the App Engine contest. After I made it through Round 1 (deploying a Fibonacci JSON-RPC service), I was presented with the Round II challenge: make an app that uses one of their more advanced APIs - blobstore, task queue, pipeline, mapper. I knew that I could use task queues in a couple different ways in Similarity Web, so at 5pm on that Friday, I started remaking it from scratch and at noon on Saturday, I submitted the fully functional remake.

Much has changed in web development in the last 5 years - both in what I personally use and what the world uses - so I thought it'd be interesting to compare the technology stack for each version of the Similarity Web.

Hosting	LiquidWeb (shared hosting)	App Engine
Backend Language	PHP	Python
Amazon API	Raw HTTP calls	Bottlenose wrapper library
Data Storage	Filesystem	Non-relational Datastore
Data Format	XML	JSON
Frontend Language	Pure JS	JS + jQuery
Visualization	Flash (AS2)	JavaScript InfoVis Toolkit

The general trend? The app is built on more scalable infrastructure, more web-friendly formats and standards, and more pre-existing libraries; it was faster to develop and works in more environments than before. Web development is getting better and better. Oh, and yes, I did get a Google I/O ticket! :)