Keegan Poppen and William Rowan

Our submission is probably a little different from most cs248 final project
submissions. Given the hubub over HTML5 and the recent stabilization of new
web standards that enable 3D applications in the browser, we decied that
the most exciting project we could work on would be based on WebGL and run
on the web.

WebGL is Javascript API for OpenGL ES in the browser. Working in this context
was both limiting and enabling. Because our game was running in the browser
we had access to a lot of convienient browser based technology such as the
native image loading functionality that has always been a part of the web.
Loading a texture is as easy as creating an image node in the DOM. Similarly,
the new HTML5 audio element gave us easy sound asset loading.

This convienience, however, was more than offset by the difficultly of finding
anything like the wealth of libraries and tools for producing and manipulating
models and animations for the desktop bound world. There are a number of pilot
webgl library projects that vary from full fledged rendering engines to
basic sets of convienience functions for compiling shaders etc. We first tried
going it alone, found it very difficult, then evaluated several different
frameworks, found that more diffiuclt, and then finally decided on a hybrid
approach using a library called SpiderGL.

SpiderGL aims to provide convinient tools for building WebGL programs without
attempting to hide the underlying implementation so we were as free to do
things ourselves where convinient. We ended up making extensive use of
SpiderGL's great matrix stack implementation and basic mesh rendering tools.
On the other hand, we ended up managing our shaders direclty and developed
our own convienience functions for compiling shader programs and setting
shader attributes and uniforms. Much of this functionality, you'll note,
was provided to us in C++ form as starter code but not available to us
on our choosen platform.

Our project was much more ambitious than just implementing a regular game
in a browser. We really wanted to do something that fully explored the space
of mobility and accessibility that the web was already providing us. Our
main motivation was to create something participatory, that would be as easy
to play as downloading a webpage, and as social as the web by taking
advantage of the fundamentally networked nature of the platform. Therefore
we deceided to produce a second component of the game that would allow
players to controll their charater using their smart phone using the
gyroscope api bulit into Mobile Webkit and the iPhone4.

This presented us with our biggest set of challenges, especially in the space
of maintaining realtime performance over wireless networking. In particular,
using the often crowded wifi hotspots at Stanford.

The most important technology for this is yet another HTML5 addition,
websockets, which implment a two-directional socket api for use by webapps
which hitherto have been limited by the origional client-server model of the
web and the only slightly more enabling technology of the xmlhttprequest.
In order to get the performance we needed, we spent a lot of time thinking
about how we could minimize our wireless traffic, especially when we had
a lot of clients all trying to connec to the same wireless access point in
order to all try and maintain a 1-to-1 mapping between their smart phone
and the model on screen.

To sum it all up, here is a list of advanced features we think we implemented:

alternate input (from the iPhone's gryoscope)
mobile devices (see previous)
realtime multiplayer over the wireles mobile web
doing anything in webgl should count
generalized replayability for awesome maneuvers you'd like to remember
hardware particle acceleration
purely procedurally generated content ;)