There are a number of alternatives available:
PyAmazon, originally written by Mark Pilgrim, then taken over by Michael Josephson. Development seems to have stalled, with the last release in August 2004.
Kung Xi’s pyaws forked pyamazon to support the then most recent Amazon Web Service and give developers more control of the incoming data. Sometime after version 0.2.0, development over at sourceforge was dropped without warning and continued at http://trac2.assembla.com/pyaws with version 0.3.0, which was released in May 2008.
This module seems to be the most widely used. It hasn’t been updated however in quite some time. A fork of this project is maintained here.
In October 2008 David Jane started pyecs after stumbling accross pyamazon. He decided that “a new, more class and iterator-oriented approach would be better.” However, it only supports a subset. Last commit was in November 2008.
There is a clever hack using boto to create the URL, although this library is originally designed to allow communication with Amazon’s cloud APIs.
So why write your own then? First and foremost, since August 15, 2009 all calls to Amazon’s Product Advertising API must be authenticated using request signatures. The afore mentioned libraries did not support this out of the box at the time. And yes... writing something from scratch is always more appealing.
More recently I stumbled across another alternative:
Dan Loewenherz’s bottlenose makes sending requests to Amazon as easy as
import bottlenose
amazon = bottlenose.Amazon("access_key_id", "secret_access_key")
response = amazon.ItemSearch(ItemId="0596520999", ResponseGroup="Images",
SearchIndex="Books", IdType="ISBN")
It has a straight-forward API, is easy to use and supports all operations out of the box. You only have to take care of processing the response. I must steal some ideas from this module!