I just saw this, Ben Strong's observation of how some websites are playing with TCP slow-start's initial window size to get better page load times. Its good to see that some people are pushing to get the protocol changed to allow for bigger initial window sizes. An initial window size of 3 is a little small given the amount of bandwidth now avaliable and the stronger reliability that you see in the tubes these days.
Its been a distressingly long time since I last blogged, but thats OK because I'm the only one that cares. My numerous lamentations now aired out, we move onto the good stuff (or bad stuff, depending on your point of view...)
So this last quarter, I found my digital design course consuming an inordinate amount of time, and with each passing week the amount of time I spent in lab per day increased at a logarithmic rate (logarthmic, because there are only so many hours a person can spend in lab before there simply ceases to be hours in the day). The premise of the class was quite simple: spend the first few weeks going through cookie-cutter labs whilst learning the ins-and-outs of the tools and VHDL, and then spend the rest of the quarter builidng your own personal project on a FPGA. My group of three proposed to create a system where a user could drive an iRobot Create by mimicking driving motions in front of a camera, an idea perhaps less ambitious then other groups in the class but nonetheless quite daunting. Unlike Microsoft's super-amazing-funtastic-galactic-awesome Kinect, our system made no use of fancy infrared projectors. Instead, we simply required the user to wear blue gloves or grip a stick tipped in blue, and place their foot over a piece of green cloth. Take a look at our demo:
You'll notice also, that on the disply we drew solid bounding boxes over the blue and green objects in the frame.
This was done almost entirely on the FPGA and was written in VHDL, with only communication over serial port handled by the PPC405 CPU. So yeah, it took a really long time.