Project Wonderland WPI MQP

Day 9

After work on Thursday, Gerard injured his foot playing softball, so he couldn’t participate in Friday’s MIT field trip (more about that in a second.) Gerard is also working from home today and tomorrow. All of us are wishing Gerard a speedy and complete recovery!

On Friday, Josh, Nicole, and Jordan ventured to MIT to attend a talk by Li-Te Chang and Steve Rohall of IBM on Bluegrass, their virtual world designed to support collaborative software development. After arriving in Cambridge, we met up with Chris Schmandt of the Speech + Mobility Group at MIT Media Lab, and bought lunch from the food trucks. We then headed over to the media lab and inspected and played around with a few of the projects they’re working on in there. Neat stuff.

Then came the talk on Bluegrass. Essentially, Bluegrass is a plugin for IBM Jazz that has an extremely similar purpose to the overall vision for our MQP. Bluegrass integrates tightly with Jazz features to provide a 3D environment that attempts to unify the higher-level aspects of working in a software development team, such as enabling team members to easily notify each other of what they’re working on, making design decisions via meetings, etc. When a software devleopment team is comprised of 120 people (the scale mentioned in the talk,) the team is naturally broken up into subteams in real life, and this is reflected in Bluegrass.

Bluegrass employs an interesting ‘forest in a basin’ metaphor. Each team gets an area in the virtual forest complete with a large tree and gazebos for [sub]team members to post materials on, for anyone else to see. Status messages ‘bubble up’ from each subteam’s area, describing what members of that subteam are doing, so seeing what a subteam is working on is as easy as walking to their area in the forest. The ‘basin’ part of the metaphor is also interesting. Because subteams are most likely heirarchically related, the most general/’highest’ subteam in the heirarchy appears in the middle of the forest. Teams directly related to that team are arranged in-world in a circle around it, and so on and so forth.

Many more features of Bluegrass were described in the talk, too many to name here. (NOTE: Josh’s first-person opinion is coming up now.) I was impressed by the attention to detail in Bluegrass’s overall design. It seemed like the developers had a very exact purpose for Bluegrass in mind throughout the design process, and stuck to it, which contrasts with our project where we’ve hit the ground running and have made design decisions on the go. I’m sure the Bluegrass developers took a lot of time to perfect the forest metaphor and to think about how to convey what could be overwhelmingly complex information in the simplest ways possible, all while taking advantage of the 3D space. I’d also like to note that I was impressed by the very intelligent questions that some of the audience members asked the developers, although I couldn’t come up with any such questions myself.

All in all, the talk and the day in general were both great experiences, and the talk certainly provided inspiration to draw from as our Wonderland MQP progresses.

Day 10

Gerard simultaneously nursed his foot and worked from home today, attempting to merge the very large number of files created by WorldBuilder that describe our actual virtual team room. Josh put the finishing touches on the HTML viewer, and is now (for better or for worse) calling it complete. The only change that we’d potentially like to make to it at this point would be to give it a catchier name then ‘the HTML Viewer.’ We’re still tossing around ideas for names (‘webshot app’, ‘netposter app’, etc) but haven’t settled on one yet. Let us know if you have any name ideas in the blog comments!

Josh also had an insightful meeting with Nicole and Jordan this morning, regarding the next component we should build for use in our Software Engineering Team Room. We scrapped some old ideas and came up with a new and exiciting one. We won’t divulge our new idea just yet, but just the preliminary writing about/sketches of the the idea already occupies three pages of yellow legal pad paper. Stay tuned!

First Breakthrough: HTML Rendering In-World

Today, we got HTML to render natively in-world! Don’t get too excited yet, because so far, our HTML Viewer component is a slightly modified version of Nigel Simpson’s Wonderland PDF viewer. It can only display whatever web page URL is hard-coded into the source code. Still, we’re very excited about this and will be rounding out the Wonderland HTML Viewer component during the coming days.

In any case, today was a pretty busy day for us. This morning, we called Nigel to ask him some specific questions about the logistics of modifying his PDF Viewer component to display web pages. Nigel was extremely helpful, and we’ll continue to ask him quetsions as we undoubtedly run into issues while completing the HTML Viewer.

We also (obviously) spent some time refactoring Nigel’s code to get it to render HTML in-world, and we were able to construct a (very) basic example demonstrating that making an HTML viewer for Wonderland is definitely possible.

The Cobra HTML rendering engine we’re using to accomplish this is released under an LGPL license, but it makes use of Mozilla’s Rhino ‘JavaScript for Java’ library for its JavaScript functionality. Rhino is released under a dual MPL 1.1/GPL 2.0 license. Amusingly, Rhino includes a library from Sun that it uses for debugging. We discussed these various licenses with Jordan, and how they might interact or potentially conflict with Wonderland’s own dual license. Jordan said that Cobra would have to undergo a license review by Sun’s legal department before it would be allowed to officially be included in Wonderland, but this is understandable.

Other things that happened today included a department “Tea” at 3:00, and we’ve scheduled weekly meetings (that will happen either in Wonderland or in person) with our WPI advisor, Gary Pollice. We forgot to mention that we got our badges yesterday! We feel official now.

We started the day by reading through Nigel’s notes on Wonderland’s PDF viewer, in order to prepare for our meeting with him tomorrow.

We then talked with Jordan and John about the method our HTML viewer would use to save images of web pages (to be used as textures in-world.) It seemed that the easiest way to scrape image data from Cobra would be to have Cobra render a page onscreen, convert its output to an image, and then save that image as a file. This approach is frowned upon for a number of reasons. The biggest problem is that if this code were to be theoretically integrated into Wonderland, pages would have to be rendered inside a hidden X11 session using Wonderland’s Server Master Client. This would have defeated the main purpose of creating the HTML viewer component in the first place: to ELIMINATE the act of scraping images from X11 applications.

Fortunately, after some heavy Googling, we found code on Martin Brown’s blog that significantly alleviated our fears. His code demonstrates how to make Cobra render directly to an image using headless AWT. We slightly expanded his code and added required error checking, and ended up writing a simple Java method with this signature:

public static void takeWebshot(String uri, Dimension size, File output) {}

This method uses the Cobra engine to save a PNG screenshot of a given web page. This is certainly a step in the right direction! Jordan thinks we may even be able to forgo saving PNG images, and paint AWT BufferedImages directly on objects in-world. We’ll have to continue working on the HTML viewer component and find out.

Also, today we saw a presentation from Sun’s chip design department, and got tours of two of their data centers. The presentation was very informative because none of the WPI MQP students knew too much about the SPARC architecture or about Sun’s processor line.

Tomorrow will be very busy indeed…watch out for tomorrow’s blog post!

Today was all of the WPI MQP Students’ first day at Sun.

Gerard and I have our own lab to work/collaborate in, and we each have our own Sun workstation running Ubuntu 8.04, with (much appreciated) 24.1″ LCDs.

In the morning, we met with Jordan, Nicole, and John to talk about the focus of our project. We described our overall vision of the Software Engineering Team Room to them, and then discussed a couple of ideas about what to tackle first.

We decided to start by building a single component that could be useful for both our team room project and Wonderland in general: an in-world HTML rendering engine component that doesn’t depend on an external web browser. In theory, this HTML rendering engine component could work in a similar fashion to Wonderland’s PDF viewer. This component could basically turn dynamic web pages into ‘posters’ in a team room; whenever team members enter the room, current Hudson build statistics could be displayed on a wall, etc.

We’d like to have this component make use of an EXISTING open-source HTML rendering engine, rather than write our own. We’ve played with a few engines, and so far Cobra looks like the most promising one, but further research is needed.

We spent the remainder of the day following and completing Jordan’s tutorials about creating custom cell types in Wonderland, and starting to do research for the aforementioned HTML rendering engine component.