Michal Migurski's notebook, listening post, and soapbox. Subscribe to this blog. Check out the rest of my site as well.

Mar 8, 2010 2:59am

walking papers plug

I've gotten a bunch of renewed interest in Walking Papers lately (previously: technology, launch, presentation, hacking). A few people have mentioned that it's difficult to install, which is true. The project originally depended on a few "HTTP ponies", single-purpose web servers with jobs like reading QR codes or rendering static maps. I've made some changes there, written up a proper installation guide, and added an explicit GNU public license to clarify the terms under which I've released the software.

It's been a lot of fun getting the project running on this guy, though:

That's a SheevaPlug, a 1.2GHz Linux server with 512MB of RAM that uses about 5W of power. It comes with Ubuntu 9.04 preinstalled, and when you plug it into a network it basically Just Works. The one I purchased from Ionics last week looks a bit different from the one above: instead of an integrated SD card slot and mini USB serial console, it has sort of a strange printed circuit hernia sticking out one side and a detachable plastic sidecar with the SD slot and mini USB. Less nice, but no matter - it does the right thing.

It's kind of an amazing, exciting piece of technology for a few interesting reasons. Mainly, I think it represents a sensible move for some of the disaster response uses we've seen Walking Papers applied to, specifically situations where remote internet is unavailable, power is hard to come by, and a highly portable little wall-wart could unobtrusively provide map printing and scanning services. It's definitely not a fast computer. My initial paper scanning, SIFTing and QR decoding tests showed it to be almost an order of magnitude slower than my Macbook at the difficult, CPU-intensive mathy inner loops of the decoder process. For the web and data service it's a champ, though.

In the near future, I'd be interested in deploying a complete OpenStreetMap server installation on one of these guys, which probably means getting a bootable volume onto the SD card instead of the severely limited half-GB of flash storage that comes inside the unit.

Anyway, what follows is the installation guide that you can find in the Paperwalking project.

Installing Paperwalking on Ubuntu 9


I've tested this guide on two platforms: Amazon's EC2 cloud-based computing service, and the SheevaPlug, a miniaturized ARM-based plug computing platform. SheevaPlug comes with Ubuntu 9.04, while EC2 can run Ubuntu 9.10 machine images. Machine images from Alestic.com seem to work well, and the guide below was produced using ami-19a34270 (alestic/ubuntu-9.10-karmic-base-20090623).

When setting up an EC2 instance, make sure it's in a security group that can accept both SSH and HTTP connections!

Start by logging into the server via SSH, as the root user.


There are a few packages that you will need to install: some base material, packaged for use by the offline image decoder, and packages to help run the public-facing website. During the last step below, you'll be asked to create a root MySQL password a few times, it's fine to leave this blank.

% apt-get update
% apt-get install curl vim screen tcsh sudo build-essential git-core
% apt-get intsall python-imaging python-numpy openjdk-6-jre-headless
% apt-get install libapache2-mod-php5 php5-gd php5-mysql mysql-server-5.1 php-pear

Paperwalking uses server packages from PHP's PEAR collection. The can be installed via the pear utility. Some of the packages below will throw warnings about deprecation, don't worry about those.

% pear install Crypt_HMAC HTTP_Request DB
% pear install Crypt_HMAC2 MDB2 MDB2#mysql

Apache's default configuration will need to be edited slightly. Edit the line with "DocumentRoot" to say: DocumentRoot /var/www/paperwalking/site/www, then restart Apache.

% pico /etc/apache2/sites-enabled/000-default
% apache2ctl restart

Try http://example.host in a browser.

Now, to install Paperwalking itself.

% cd /var/www
% git clone http://github.com/migurski/paperwalking.git paperwalking
% cd paperwalking
% git submodule init && git submodule update
% mysql -u root
	> create database paperwalking character set='utf8';
	> grant select, insert, update, delete, lock tables on paperwalking.* to paperwalking@localhost identified by 'w4lks';
	> flush privileges;
	> quit;

% mysql -u root paperwalking < site/doc/create.mysql
% cd site && make

The main site configuration information is kept in lib/init.php. Paperwalking comes with a blank that you have to copy and edit with your favorite text editor.

% cp lib/init.php.txt lib/init.php
% pico lib/init.php

Change these two lines (choose your own password):

  1. define('DB_DSN', 'mysql://paperwalking:w4lks@localhost/paperwalking');
  2. define('API_PASSWORD', 'swordfish');

Now try http://example.host in a browser to see it work. Make a new print, it might take a little while (we'll get back to this later). Print it, scan it, or just convert it to a JPEG, and post the image back to your instance of Paperwalking. Note that it's just sitting there, "queued for processing". Keep this browser window open, because we need to build SIFT.

% cd ../decoder
% make

Some stuff will scroll by, you may see compiler warnings, ignore these unless actual errors show up. Run the decoder once with the password chosen above.

% python poll.py -p swordfish -b http://localhost once

Watch the scan page in your browser update as it progresses through the image.

If it works, then re-run poll.py without the "once" argument at the end.

We don't have a proper daemon wrapper for this yet, so I've just been running it in screen as a cheap way to get a long-running process going. It sounds ghetto, but I've had processes like this one stay up for months at a time this way.

Use of screen to maintain long-running shell sessions is described here.

Now for a few optional niceties.

First, we can speed up the creation of print maps by using Modest Maps ws-compose.

% cd ~
% git clone http://github.com/migurski/modestmaps-py.git modestmaps

And in a screen as above:

% cd modestmaps
% python ws-compose.py

Back in the site configuration, configure ws-compose.

% pico /var/www/paperwalking/site/lib/init.php 

Finally we can get applications IDs for three external information services, for specific improvements to the site.

  • GeoPlanet is used in conjunction with the "Find" button on the map composition page.
  • Flickr does our reverse-geocoding, so that maps of arbitrary places can be given names
  • Cloudmade provides a few extra map styles.

Comments (1)

  1. Posted by Julian Bleecker on Tuesday, March 9 2010 1:55pm EST

Sorry, no new comments on old posts.

July 2017
Su M Tu W Th F Sa

Recent Entries

  1. blog all dog-eared pages: human transit
  2. the levity of serverlessness
  3. three open data projects: openstreetmap, openaddresses, and who’s on first
  4. building up redistricting data for North Carolina
  5. district plans by the hundredweight
  6. baby steps towards measuring the efficiency gap
  7. things I’ve recently learned about legislative redistricting
  8. oh no
  9. landsat satellite imagery is easy to use
  10. openstreetmap: robots, crisis, and craft mappers
  11. quoted in the news
  12. dockering address data
  13. blog all dog-eared pages: the best and the brightest
  14. five-minute geocoder for openaddresses
  15. notes on debian packaging for ubuntu
  16. guyana trip report
  17. openaddresses population comparison
  18. blog all oft-played tracks VII
  19. week 1,984: back to the map
  20. bike eleven: trek roadie