Creating a mobile guided tour application for the Android platform

The first schedule is posted!

1. A “hello world” program proving my IDE, emulator, and device are cooperating and in good order
2. A proof-of-concept with GPS and Google Maps functionality, but a dummy data set. It may not be able to tell you about the history of Cofrin Hall, but it will respond to simulated GPS tracks in the debugger. It is here I will develop my first test cases.
3. A third prototype will be made with real data (images, locations, and factoids). I anticipate architectural changes in response to new insights from the second prototype. I do not intend to supply a schedule for the third prototype until Fred Brooks has told me how much of prototype two I’m throwing away.

On a related note, open source project management software leaves much to be desired, especially to someone who had to Google “Gantt Chart.”  I anticipate prettier output from a trial of Microsoft Project.

VirtualBox provides a Remote Desktop server for connecting to, booting, and controlling its virtual machines. I found this remote desktop interminably slow, with minute long screen refreshes. However, connecting to the native Windows remote desktop on the same machine is almost as snappy as direct access.

It is infinitely faster to control the VM from the WINDOWS remote desktop than to directly connect to its OWN remote desktop. I guess Microsoft’s API acceleration really does work, and the VirtualBox Guest Additions tie into Windows really well.

A little Googling resolved a bizarre glitch with VirtualBox and processor usage.

My desktop boasts a 2.66 GHz Core 2 Duo processor with 6 GB of RAM.  VirtualBox hosting my little CentOS server and development environment would peg one CPU core regardless of load.

It seems ridiculous to have an idle VM consume 50% of my CPU cycles.  According to this page, the solution is two idle VMs.

So, I configured a second virtual machine with 4 MB of RAM, no operating system, and no fixed disk. Sure enough, my host system CPU usage dropped to around 2%. Strange.

One nice thing about a Linux development environment is package management. Anything you could possibly ever want is just a download away.

However, there are a LOT of downloads.  Kernel.  The X Windowing System and GNOME.  VM Guest Additions.  GCC and Kernel-Sources for compiling said guest additions – they’re kernel objects.  Apache, PHP, and MySQL.  WordPress.  Eclipse.  The Android SDK.  Android SDK Components and images.

In fact, I’m still downloading.  At least I didn’t need to compile much – trying to hack a newer kernel into CentOS for OpenGL support took over an hour of compile time and broke GRUB.  Hooray for VM snapshots!

Although I finally have a fully-configured IDE, “Hello World” might have to wait ’til the morrow.

I retract my previous statement about VirtualBox’s RDP server. Although VRDP was impressive over the LAN, my upstream bandwidth leaves something to be desired. It was great fun trying to remotely change Gnome’s screen resolution – the 15-second revert timer would count down before the screen could redraw, and the resolution would change back.

A few times back and forth of that and suddenly my page went offline. Remote desktop would connect and authenticate but served up a meditative quantity of inky darkness.

Turns out I crashed the virtual machine. Seems VRDP is rather finicky.

Moved journal and website over to a WordPress installation.  That’s right; I upgraded the Web from 1.0 to 2.0.

This sits on a CentOS virtual machine that will also serve as my primary development environment.  VirtualBox has a very snappy RDP server, meaning I can log into the VM and do Android development from any XP or later computer with the standard Windows Remote Desktop client.

This somewhat eliminates the need for SVN as well.  Naturally, I’m the only developer on my project, so source control won’t be quite as important.

All that remains is configuring Eclipse for use with the Android SDK.